Wildlife image compression and encrypted transmission method driven by edge computing

Abstract

The application discloses an edge computing driven wild animal image compression and encrypted transmission method, relates to the technical field of image processing, and comprises the following steps: constructing a cache access density curve based on the task triggering frequency and the cache calling times of an edge computing device within a unit time, recording the time sequence of the maximum available space and the current available space of a cache area, and extracting the cache fluctuation period and the instantaneous compression amplitude; calculating the difference between the cache allocation duration and the release delay, and constructing a cache response lag index. The application realizes quantitative perception and prediction of cache risks by constructing multi-dimensional cache state indexes and generating a cache resource pressure bearing coefficient. When the cache resource pressure bearing coefficient exceeds the overpressure risk threshold, the system automatically performs image frame adaptive reduction, relieves the load while retaining event continuity, and effectively improves the stability and resource utilization efficiency of edge devices in high-density image processing scenarios.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to a wildlife image compression and encryption transmission method driven by edge computing. Background Art

[0002] Edge computing-driven wildlife image compression and encryption transmission refers to a technical solution in which intelligent monitoring terminals deployed in the wild (such as edge computing devices or AI cameras) perform local compression and encryption on captured wildlife image data before remote transmission. Specifically, this system integrates computing-capable edge nodes on the terminal side. These nodes use target detection algorithms to identify animal targets in images. Adaptive image compression techniques (such as ROI-based local fidelity compression) preserve high resolution in key areas while downsampling background regions, significantly reducing image size. Furthermore, data encryption is performed locally using lightweight encryption algorithms (such as AES or symmetric encryption combined with blockchain hashing) to ensure image privacy and tamper-proofing during transmission. This approach not only reduces the bandwidth required for image transmission, improving system real-time performance and energy efficiency, but also enables efficient and reliable processing of sensitive ecological monitoring data without relying on cloud computing. It is particularly suitable for off-grid, low-power ecological monitoring scenarios.

[0003] The existing technology has the following deficiencies:

[0004] In existing technologies, image compression and encryption processing typically relies on edge computing nodes to schedule serial tasks in a time-division multiplexing manner to reduce device resource consumption and adapt to low-power environments. However, in field monitoring scenarios, when multiple high-speed image acquisition events (such as wildlife migration and cluster activities) are triggered continuously, it can easily put pressure on the edge device's cache resources, causing compression and encryption tasks to overlap on the timeline. Because the cache management mechanism lacks dynamic concurrent scheduling capabilities, it is very easy for the encryption module to call the compressed frame before it is fully processed, or for some encrypted data to be overwritten by new tasks. This can lead to the destruction of the image compression structure or incomplete encrypted content. The resulting image file cannot be restored during the decoding phase, causing the image data to fail in batches. This type of problem is particularly prominent in high-density event perception scenarios, seriously affecting the logical continuity of image data and the stability of system operation.

[0005] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0006] The purpose of the present invention is to provide an edge computing-driven wildlife image compression and encryption transmission method. By constructing a cache access density curve, a cache response lag index, and a load increase rate, a comprehensive perception of the dynamic usage status of cache resources is achieved, and a normalized cache resource pressure coefficient is generated by integration to quantify the degree of tension of cache resources and their potential failure risk. When the cache resource pressure coefficient exceeds the preset risk threshold, the system automatically enters a high-risk processing state and, based on the current resource pressure level, performs an adaptive uniform deletion strategy on a set of high-frequency triggered image frames, effectively alleviating the cache load while preserving the temporal continuity of events and the integrity of image expression as much as possible, thus achieving a shift from the traditional "passive response to cache overflow" to "active prediction and regulation of cache risks" to solve the problems in the above-mentioned background technology.

[0007] To achieve the above objectives, the present invention provides the following technical solution: an edge computing-driven wildlife image compression and encryption transmission method, comprising the following steps:

[0008] S101. Construct a cache access density curve based on the task trigger frequency and cache call count per unit time of the edge computing device. Record the time series of the maximum available space and the current available space in the cache area. Extract the cache fluctuation period and instantaneous compression amplitude to reflect the dynamic usage characteristics of the cache resources.

[0009] S102. During each image compression and encryption process, calculate the difference between the cache allocation duration and the release delay to construct a cache response hysteresis index; perform a weighted fusion of the cache response hysteresis index and the cache access density curve to generate an intermediate feature vector representing the continuity of cache resource occupancy and the degree of release imbalance;

[0010] S103: Collect the intensity change trend of the image trigger event and the average accumulation time of the corresponding processing task, calculate the load increase rate of the cache processing, and introduce a task accumulation saturation function. The task accumulation saturation function is a Sigmoid function. It performs nonlinear amplification mapping on the cache resource pressure under the sudden event and identifies the cache limit approach behavior;

[0011] S104: The cache access density curve, the intermediate feature vector, and the load increase rate are integrated to construct a multi-dimensional pressure state vector; the pressure state vector is mapped into a cache resource pressure coefficient ranging from 0 to 1 through normalization processing, which is used to characterize the intensity of cache resource usage and potential failure risk of the edge computing device in high-frequency image processing scenarios;

[0012] S105. Compare the cache resource pressure coefficient with the preset overvoltage risk threshold. When the pressure index exceeds the overvoltage risk threshold, determine that the edge device has entered a high-risk state for cache resources. In the high-risk state for cache resources, based on the current cache resource pressure level of the edge device, adaptively and uniformly delete the image frame sets of the same continuous event that are continuously triggered within a short period of time to reduce cache pressure and prevent batch failure of image data due to cache conflicts.

[0013] Preferably, the steps of constructing a cache access density curve and extracting cache resource dynamic usage characteristics further include:

[0014] Based on the number of image trigger requests received per second by the image processing module and the number of compression or encryption processes actually initiated, a corresponding relationship between task trigger frequency and cache call frequency is established. The cache call density per unit time is calculated using a sliding time window to form a preliminary cache access density curve.

[0015] During cache resource usage, the maximum available space and the current remaining available space in the cache area are continuously monitored to generate time series data. The periodic fluctuation characteristics of cache space changes are extracted through Fourier transform or wavelet analysis to determine cache resource usage trends and sudden change patterns.

[0016] Combining the cache call density change slope and cache space fluctuation gradient, the instantaneous compression amplitude of cache resources per unit time is calculated, and the degree of cache burst contraction is determined based on the compression amplitude. This allows the dynamic usage characteristics of cache resources to be obtained in real time in the image processing link, improving the early warning of task scheduling and system stability.

[0017] Preferably, the steps of constructing a cache response hysteresis indicator and generating an intermediate feature vector are:

[0018] In each image compression and encryption process, the time point when the current task applies for cache resources and the time point when the corresponding cache resources are released for the task are recorded respectively, and the time difference between the two is calculated to obtain the cache response lag corresponding to the task;

[0019] Normalize the cache response lag with the cache access density curve within the task's time period and perform a weighted combination based on a preset fusion weight factor to appropriately amplify the weight impact of lag behavior in processing scenarios with high task density.

[0020] Based on the cache response lag index generated during the continuous processing of multiple tasks, we statistically obtained three dimensional indicators: lag peak, average lag level, and fluctuation amplitude. These indicators serve as basic parameters for measuring cache usage continuity and resource recovery efficiency.

[0021] The three-dimensional hysteresis statistical features are fused and constructed into a multi-dimensional intermediate feature vector to characterize the balance of cache resource usage and release hysteresis distribution characteristics of edge computing devices under image intensive processing state, thereby enhancing the accuracy and robustness of cache pressure status assessment.

[0022] Preferably, the steps of calculating the cache load increase rate and extracting the cache limit approach behavior include:

[0023] Continuously monitor the intensity trend of image acquisition trigger events, construct an event intensity curve based on the number and frequency of trigger events per unit time, and simultaneously record the average waiting time of each image processing task from the time it enters the processing queue to the time it is scheduled for execution, to dynamically characterize the current task backlog and processing queue pressure;

[0024] The cache processing load growth rate is calculated based on the product of the gradient of the event intensity curve and the growth rate of the average task accumulation time. This load growth rate is used to reflect the increasing speed of cache scheduling pressure on edge computing devices in high-frequency image input scenarios.

[0025] A task accumulation saturation function with nonlinear mapping features is constructed. A Sigmoid function, exponential curve, or other monotonically increasing function is used to weightedly amplify the load increase rate. This allows cache pressure to increase exponentially when sudden image events are frequently triggered, thereby improving the system's sensitivity to high-intensity event shocks.

[0026] The nonlinear load increase value after the above processing is used as the input feature, combined with the usage boundary of the cache resources within the time window, to identify the extreme approach behavior of the cache resources reaching the full load state. This is used in the subsequent pressure state modeling and risk prediction mechanism to achieve high-precision response of edge devices to sudden image processing pressure.

[0027] Preferably, the steps of constructing a multi-dimensional pressure state vector and generating a cache resource pressure coefficient include:

[0028] The cache access density curve representing cache access frequency, the intermediate feature vector representing cache response lag and release imbalance, and the load growth rate reflecting the dynamic growth trend of image processing pressure are aligned and synchronized with the time window to construct a multidimensional feature dataset at a unified time scale. This ensures temporal consistency and statistical stability when integrating different types of indicators.

[0029] Using weighted linear combination, principal component analysis (PCA), or other multi-dimensional dimensionality reduction and fusion methods, the aforementioned feature datasets are fused to generate a pressure state vector that comprehensively reflects the degree of cache resource usage. Combined with a preset dynamic weight adjustment mechanism, this gives higher computational weights to feature dimensions that are more sensitive to system impact in different operating scenarios, thereby enhancing the system's ability to identify early signs of cache failure.

[0030] The pressure state vector is normalized and mapped to a cache resource pressure coefficient ranging from 0 to 1. The closer the cache resource pressure coefficient is to 1, the tighter the cache resource usage is and the closer the edge computing device is to the upper critical limit of the cache processing capacity.

[0031] The cache resource pressure coefficient is used as a reference indicator for quantitative risk measurement in system operation, providing a basis for subsequent execution of the image deletion mechanism, and realizing real-time perception and active response to cache risks in the image processing link.

[0032] Preferably, in a high-risk state for cache resources, adaptive uniform deletion is performed on a set of image frames of the same continuous event that is continuously triggered within a short period of time. The specific steps are as follows:

[0033] When the cache resource enters a high-risk state, the image deletion ratio coefficient that needs to be deleted in the image frame set is calculated based on the current cache resource pressure coefficient and the preset overpressure risk threshold. The calculation formula is as follows: ,in: The cache resource pressure coefficient indicates the current cache pressure intensity. The larger the value, the tighter the resource. ; is the overvoltage risk threshold of cache resources. The deletion mechanism is triggered when is the deletion sensitivity adjustment factor, which is used to control the nonlinear response speed of the deletion intensity. ; The image deletion ratio coefficient indicates the ratio that needs to be deleted from the image frame set. The larger the value, the more deletions are made.

[0034] Suppose the set of image frames of the same event that are triggered continuously in a short period of time is ,in N is the original frame number, and the image deletion ratio coefficient calculated according to step 1 , determine the number of frames that should be retained, and the expression for determining the number of frames that should be retained is: ,in: K The number of image frames to be retained is reduced from the original number of frames according to the deletion ratio N Calculation in Indicates rounding up;

[0035] Then, keep the K Frame images, forming a set of image frames retained after deletion (The image frame set retains frames extracted at equal intervals to cover the integrity of the original event as much as possible). ,in: ,in: The first one retained after deletion Frame image; Indicates the index of the image frame retained after deletion, .

[0036] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0037] The present invention realizes a comprehensive perception of the dynamic usage status of cache resources by constructing a cache access density curve, a cache response lag index, and a load increase rate, and integrates and generates a normalized cache resource pressure coefficient, which is used to quantitatively determine the tension of cache resources and their potential failure risks. When the cache resource pressure coefficient exceeds the preset risk threshold, the system automatically enters a high-risk processing state, and according to the current resource pressure level, performs an adaptive uniform deletion strategy on the high-frequency triggered image frame set, which effectively alleviates the cache load and preserves the temporal continuity of events and the integrity of image expression as much as possible, realizing the transformation from the traditional "passive response to cache overflow" to "active prediction and regulation of cache risks", significantly improving the system stability, image data integrity and cache resource utilization efficiency of edge devices in high-density image processing scenarios, and has good engineering adaptability and promotion and application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction to the drawings required for use in the embodiments will be given below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0039] Figure 1 This is a flow chart of the method for compression and encryption transmission of wildlife images driven by edge computing of the present invention. DETAILED DESCRIPTION

[0040] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of this disclosure will be thorough and complete and will fully convey the concepts of the example embodiments to those skilled in the art.

[0041] The present invention provides Figure 1The edge computing-driven wildlife image compression and encryption transmission method shown includes the following steps:

[0042] S101. Construct a cache access density curve based on the task trigger frequency and cache call count per unit time of the edge computing device. Record the time series of the maximum available space and the current available space in the cache area. Extract the cache fluctuation period and instantaneous compression amplitude to reflect the dynamic usage characteristics of the cache resources.

[0043] The steps of constructing a cache access density curve and extracting cache resource dynamic usage characteristics further include:

[0044] Based on the number of image trigger requests received per second by the image processing module and the number of compression or encryption processes actually initiated, a corresponding relationship between task trigger frequency and cache call frequency is established. The cache call density per unit time is calculated using a sliding time window to form a preliminary cache access density curve.

[0045] During cache resource usage, the maximum available space and the current remaining available space in the cache area are continuously monitored to generate time series data. The periodic fluctuation characteristics of cache space changes are extracted through Fourier transform or wavelet analysis to determine cache resource usage trends and sudden change patterns.

[0046] Combining the cache call density change slope and cache space fluctuation gradient, the instantaneous compression amplitude of cache resources per unit time is calculated, and the degree of cache burst contraction is determined based on the compression amplitude. This allows the dynamic usage characteristics of cache resources to be obtained in real time in the image processing link, improving the early warning of task scheduling and system stability.

[0047] By quantifying and dynamically modeling the cache resource usage in edge computing devices in real time, a "cache access density curve" is constructed to reflect the relationship between image processing task density and cache call intensity. Combined with time-series changes in cache space, this curve comprehensively characterizes the actual cache resource usage and pressure trends per unit time. Traditional image processing systems often use static thresholds or fixed priority strategies to schedule compression and encryption tasks. This lacks accurate awareness of dynamic fluctuations in cache resources and is prone to scheduling imbalances, data conflicts, and cache overflows in high-frequency triggering scenarios. This process not only captures the periodic fluctuations in the cache caused by task accumulation during continuous image processing, but also quantifies the magnitude of cache space compression over short periods of time, thereby forming an early warning indicator of system load. This provides a basis for subsequent resource scheduling optimization and task removal decisions, enabling edge devices to achieve greater task scheduling flexibility and fault tolerance in low-power, high-data-density field monitoring environments, fundamentally improving system stability and image data processing integrity.

[0048] S102. During each image compression and encryption process, calculate the difference between the cache allocation duration and the release delay to construct a cache response hysteresis index; perform a weighted fusion of the cache response hysteresis index and the cache access density curve to generate an intermediate feature vector representing the continuity of cache resource occupancy and the degree of release imbalance;

[0049] The steps to construct the cache response lag indicator and generate the intermediate feature vector are:

[0050] In each image compression and encryption process, the time point when the current task applies for cache resources and the time point when the corresponding cache resources are released for the task are recorded respectively, and the time difference between the two is calculated to obtain the cache response lag corresponding to the task;

[0051] Normalize the cache response lag with the cache access density curve within the task's time period and perform a weighted combination based on a preset fusion weight factor to appropriately amplify the weight impact of lag behavior in processing scenarios with high task density.

[0052] Based on the cache response lag index generated during the continuous processing of multiple tasks, we statistically obtained three dimensional indicators: lag peak, average lag level, and fluctuation amplitude. These indicators serve as basic parameters for measuring cache usage continuity and resource recovery efficiency.

[0053] The three-dimensional hysteresis statistical features are fused and constructed into a multi-dimensional intermediate feature vector to characterize the balance of cache resource usage and release hysteresis distribution characteristics of edge computing devices under image intensive processing state, thereby enhancing the accuracy and robustness of cache pressure status assessment.

[0054] By quantifying the timeliness of cache resource usage by image processing tasks in edge devices, a dynamic "cache response lag index" is established and integrated with the system's current cache access density curve to accurately assess the continuous occupancy characteristics of cache resources and the imbalance in release latency. Traditional image processing systems typically only focus on whether a task is successfully completed, while ignoring the actual utilization efficiency of cache resources during task execution. This is especially true in edge computing scenarios where tasks are frequently and densely triggered. If cache allocation is delayed or released inefficiently, subsequent tasks will encounter waiting times or even overwrites when calling the cache, seriously affecting the continuity and reliability of the entire processing chain.

[0055] By recording the "resource allocation moment" and "resource release completion moment" of each image compression and encryption task at the cache level and calculating the difference between the two durations, we can quantify the duration of cache occupancy for that task, thereby constructing a metric that reflects the resource utilization efficiency of individual tasks: cache response lag. This metric itself is capable of identifying lagging tasks and monitoring system congestion trends, but in high-concurrency scenarios, the significance of a single lag value is diluted. Therefore, this method further combines lag with the cache access density curve through a weighted fusion. This amplifies the weight of lag behavior when tasks are heavily backlogged and cache calls are frequent, thereby more sensitively reflecting the system risks that may arise from untimely resource release.

[0056] Ultimately, the "intermediate feature vector" generated through this fusion process comprehensively represents the continuity, lag, and release balance of cache resources across multiple tasks, serving as a key input variable for subsequent systems to determine cache pressure and trigger scheduling adjustments or image pruning strategies. This approach not only improves the resolution of cache resource status assessment but also provides edge devices with proactive adjustment capabilities based on behavioral feature feedback, significantly enhancing the system's stability and robustness during peak image processing periods.

[0057] S103: Collect the intensity change trend of the image trigger event and the average accumulation time of the corresponding processing task, calculate the load increase rate of the cache processing, and introduce the task accumulation saturation function to perform nonlinear amplification mapping on the cache resource pressure under the sudden event to identify the cache limit approach behavior;

[0058] The steps to calculate the cache load increase rate and extract the cache limit approaching behavior include:

[0059] Continuously monitor the intensity trend of image acquisition trigger events, construct an event intensity curve based on the number and frequency of trigger events per unit time, and simultaneously record the average waiting time of each image processing task from the time it enters the processing queue to the time it is scheduled for execution, to dynamically characterize the current task backlog and processing queue pressure;

[0060] The cache processing load growth rate is calculated based on the product of the gradient of the event intensity curve and the growth rate of the average task accumulation time. This load growth rate is used to reflect the increasing speed of cache scheduling pressure on edge computing devices in high-frequency image input scenarios.

[0061] A task accumulation saturation function with nonlinear mapping features is constructed. A Sigmoid function, exponential curve, or other monotonically increasing function is used to weightedly amplify the load increase rate. This allows cache pressure to increase exponentially when sudden image events are frequently triggered, thereby improving the system's sensitivity to high-intensity event shocks.

[0062] The nonlinear load increase value after the above processing is used as the input feature, combined with the usage boundary of the cache resources within the time window, to identify the extreme approach behavior of the cache resources reaching the full load state. This is used in the subsequent pressure state modeling and risk prediction mechanism to achieve high-precision response of edge devices to sudden image processing pressure.

[0063] To identify "limit-approaching behavior" indicating that cache resources are nearing capacity, the system first uses the load increase rate, processed by a nonlinear function, as a core input feature. This load increase rate characterizes the rate of increase in system stress under the accumulation of image processing tasks. Next, cache resource usage boundary data is collected over a continuous time window, including metrics such as cache space used per unit time, remaining available space, and the gradient of the utilization rate change. By constructing a joint surface of cache utilization rate and time, the system determines whether the cache resources are continuously approaching their upper limit. Specifically, when the nonlinear load increase value is high (e.g., above 0.8) and the cache utilization rate shows a monotonically increasing trend within the sliding time window, and the growth slope exceeds a set threshold, the system is considered to have entered the cache resource limit-approaching range. Furthermore, by setting multiple overload critical zones (e.g., warning zones, alert zones, and critical zones) and combining the results of comparing cache release rate with the frequency of new task entry, the accuracy of this judgment can be further improved. When these combined conditions are met, the system determines that the cache is in a state of "approaching limit behavior," triggering strategies such as image task pruning, cache reallocation, or processing throttling to proactively mitigate cache overflow risks. This mechanism, with its high dynamics and predictive capabilities, is a crucial technical step in ensuring cache stability on edge devices.

[0064] The purpose of this step is to provide edge computing devices with a dynamic cache load assessment mechanism based on the collaborative modeling of image trigger event intensity and task accumulation characteristics, so as to achieve nonlinear enhanced recognition of cache resource pressure changes and early prediction of "limit approach state". Image monitoring systems deployed in field environments face highly uncertain trigger event intensity, such as group animal migration and sudden cluster activities. Such events often generate a large amount of image data in a short period of time, causing edge devices to bear sudden high-density task loads. If the system's current cache resource pressure cannot be assessed in real time and its state close to the collapse boundary cannot be identified, serious consequences such as image data processing failure, cache overflow, and bulk data loss may occur.

[0065] By continuously capturing the intensity trend of image-triggered events and combining it with the average waiting time of image processing tasks after entering the processing queue, the system's current task backlog is reflected. Furthermore, the "cache load growth rate" is calculated using the slope of the event intensity change and the accumulation time growth rate. This metric captures the growth trend of system processing pressure per unit time. However, in real-world scenarios, the impact of pressure often develops non-linearly. Especially during sudden, high-intensity events, critical load conditions often rapidly approach resource bottlenecks in an exponential or saturation manner. Therefore, a "task accumulation saturation function" is introduced. Using a sigmoid function or exponential growth model, the load growth rate is nonlinearly amplified to more sensitively identify "limiting behavior"—when cache resources are approaching capacity limits. This identification mechanism significantly improves the system's perception of cache failure risks in high-pressure scenarios, enabling the device to take proactive countermeasures (such as task pruning and delayed scheduling) before the load becomes truly out of control. This mechanism has significant engineering value and ensures system robustness, and is a key link in ensuring the continuity and data integrity of the image processing chain.

[0066] S104: The cache access density curve, the intermediate feature vector, and the load increase rate are integrated to construct a multi-dimensional pressure state vector; the pressure state vector is mapped into a cache resource pressure coefficient ranging from 0 to 1 through normalization processing, which is used to characterize the intensity of cache resource usage and potential failure risk of the edge computing device in high-frequency image processing scenarios;

[0067] The steps of constructing a multi-dimensional pressure state vector and generating a cache resource pressure coefficient include:

[0068] The cache access density curve representing cache access frequency, the intermediate feature vector representing cache response lag and release imbalance, and the load growth rate reflecting the dynamic growth trend of image processing pressure are aligned and synchronized with the time window to construct a multidimensional feature dataset at a unified time scale. This ensures temporal consistency and statistical stability when integrating different types of indicators.

[0069] Using weighted linear combination, principal component analysis (PCA), or other multi-dimensional dimensionality reduction and fusion methods, the aforementioned feature datasets are fused to generate a pressure state vector that comprehensively reflects the degree of cache resource usage. Combined with a preset dynamic weight adjustment mechanism, this gives higher computational weights to feature dimensions that are more sensitive to system impact in different operating scenarios, thereby enhancing the system's ability to identify early signs of cache failure.

[0070] When fusing multidimensional feature datasets representing cache resource status (including cache access density curves, intermediate feature vectors, and cache load growth rates), principal component analysis (PCA) is an efficient and interpretable dimensionality reduction method that can compress redundant feature dimensions while retaining key information, improving computing efficiency and response speed in edge computing environments. The specific process is as follows:

[0071] First, preprocess the original feature dataset. Because the dimensions of each feature are different—for example, cache access density is measured in accesses per second, intermediate feature vectors are dimensionless statistics, and the load increase rate is a ratio-based time-varying indicator—a unified standard is necessary. Z-score normalization is typically used. This involves subtracting the mean from each feature vector and dividing it by its standard deviation. This transforms all features into zero mean and unit variance, eliminating the influence of dimensional differences on the principal component analysis results.

[0072] Next, we construct a covariance matrix based on the standardized feature matrix. The covariance matrix describes the linear correlation between different features. Essentially, it measures whether any two features have a consistent growth or decrease trend in the sample data. For example, if the cache load growth rate is highly positively correlated with the intermediate lag feature, it indicates that resource release is likely to lag when system pressure increases.

[0073] Next, we perform eigenvalue decomposition on the covariance matrix, obtaining sets of eigenvalues ​​and their corresponding eigenvectors. Each eigenvalue represents the proportion of variance explained by that principal component (i.e., the direction of the eigenvector) in the original data, while the eigenvector itself represents a new principal component direction. After sorting these eigenvalues ​​from largest to smallest, we select the top k principal components based on their cumulative contribution (e.g., achieving 90% information retention), preserving the most information with the fewest dimensions.

[0074] The original data is then linearly projected using the selected principal component vectors, mapping the original high-dimensional features into a newly selected k-dimensional subspace. This generates a new vector, called the pressure state vector. This vector comprehensively expresses the most important trends in the original features, significantly reducing dimensionality while retaining critical system operating status information.

[0075] Finally, the resulting pressure state vector, a compact representation of cache resource usage, can be further normalized or mapped to a risk function to generate a cache pressure index between 0 and 1. This index not only reflects the current level of resource pressure on the system but also serves as a core criterion for subsequent modules such as image pruning and adaptive scheduling. By introducing PCA, the system can maintain efficient, stable, and risk-aware operation while handling large-scale image-triggered tasks.

[0076] The pressure state vector is normalized and mapped to a cache resource pressure coefficient ranging from 0 to 1. The closer the cache resource pressure coefficient is to 1, the tighter the cache resource usage is and the closer the edge computing device is to the upper critical limit of the cache processing capacity.

[0077] The cache resource pressure coefficient is used as a reference indicator for quantitative risk measurement in system operation, providing a basis for subsequent execution of the image deletion mechanism, and realizing real-time perception and active response to cache risks in the image processing link.

[0078] By integrating and modeling multi-source heterogeneous indicators, a cache resource pressure coefficient is constructed that can dynamically quantify and uniformly measure system cache pressure, thereby providing accurate operational status assessment and risk identification for edge computing devices in high-frequency image processing scenarios. During the image compression and encryption tasks performed by edge devices, cache resource occupancy is a dynamic process affected by multiple factors. A single indicator (such as the number of tasks or cache utilization) often cannot accurately capture the full picture of the cache failure risk faced by the device. Therefore, this step proposes to comprehensively model three types of key indicators, including the cache access density curve, the intermediate eigenvector (reflecting the continuity and release lag of the cache), and the load increase rate (reflecting the accumulation trend of sudden tasks).

[0079] First, by aligning the time and space of these three types of features, a multidimensional pressure state vector is constructed at a unified time scale, resolving the issues of indicator asynchrony and scale inconsistency. Second, weighted fusion and principal component analysis are used to aggregate the high-dimensional feature data into a representative comprehensive vector. This allows the system to automatically adjust weights based on actual risk sensitivity under different operating conditions, improving its ability to predict potential failures. Finally, through normalization, this state vector is mapped to a cache resource pressure coefficient in the interval [0, 1]. The closer the value of this cache resource pressure coefficient is to 1, the more limited the cache resources are and the closer the device is to failure.

[0080] The greatest value of the cache resource pressure coefficient lies in that it converts multiple complex time series features into an interpretable, traceable, and driveable quantitative indicator. It can serve as the core trigger condition for internal task scheduling optimization, image deletion strategy activation, risk notification, and alarm mechanism in the system, enabling edge computing devices to have highly intelligent load self-perception and risk active intervention capabilities in complex outdoor environments, significantly improving the system's stability, robustness, and task processing continuity.

[0081] S105. Compare the cache resource pressure coefficient with a preset overpressure risk threshold. When the pressure index exceeds the overpressure risk threshold, determine that the edge device has entered a cache resource high-risk state. In the cache resource high-risk state, based on the current cache resource pressure level of the edge device, adaptively and uniformly delete image frames of the same continuous event that are continuously triggered within a short period of time to reduce cache pressure and prevent batch invalidation of image data due to cache conflicts.

[0082] When cache resources are at high risk, adaptively and evenly delete the image frames of the same continuous event that is triggered continuously within a short period of time. The specific steps are as follows:

[0083] When the cache resource enters a high-risk state, the image deletion ratio coefficient that needs to be deleted in the image frame set is calculated based on the current cache resource pressure coefficient and the preset overpressure risk threshold. The calculation formula is as follows: ,in: The cache resource pressure coefficient indicates the current cache pressure intensity. The larger the value, the tighter the resource. ; is the overvoltage risk threshold of cache resources. The deletion mechanism is triggered when is the deletion sensitivity adjustment factor, which is used to control the nonlinear response speed of the deletion intensity. ; The image deletion ratio coefficient indicates the ratio that needs to be deleted from the image frame set. The larger the value, the more deletions are made.

[0084] The purpose of this step is to convert the degree of cache overpressure into a measurable deletion decision variable, enabling the system to intelligently adjust the image processing load based on the current resource pressure, avoiding the extreme strategies of "full deletion" or "full retention", and improving processing accuracy and cache release efficiency.

[0085] Suppose the set of image frames of the same event that are triggered continuously in a short period of time is ,in N is the original frame number, and the image deletion ratio coefficient calculated according to step 1 , determine the number of frames that should be retained, and the expression for determining the number of frames that should be retained is: ,in: K The number of image frames to be retained is reduced from the original number of frames according to the deletion ratio N Calculation in Indicates rounding up;

[0086] Then, keep the K Frame images, forming a set of image frames retained after deletion (The image frame set retains frames extracted at equal intervals to cover the integrity of the original event as much as possible). ,in: ,in: The first one retained after deletion Frame image; Indicates the index of the image frame retained after deletion, ;

[0087] The purpose of this formula is to select the image frames at equal intervals from a large number of image frames that are generated at different intervals and in a burst. K frames to maximize the temporal coverage of events.

[0088] For example, if:

[0089] There are a total of N=100 image frames that need to be retained K = 10 pictures, hope from the beginning to the end, every Keep one frame.

[0090] The expression This is the same proportional mapping extraction, but by "rounding up" to ensure that the frame number is a legal integer to prevent decimal frame numbers.

[0091] The purpose of this step is to apply the deletion ratio decision to the actual image sequence and ensure the temporal continuity and spatial representativeness of event expression by extracting frames at equal intervals. At the same time, it greatly reduces memory usage and scheduling pressure, alleviates cache resource bottlenecks while ensuring the validity of ecological monitoring data, and improves the overall stability and response elasticity of the system.

[0092] The purpose of this step is to build an edge image task load reduction mechanism with real-time feedback and adjustment capabilities by comparing the threshold of the cache resource pressure coefficient and linking it with the image adaptive deletion mechanism. This allows for precise intervention when the cache resources are at risk of overload, ensuring continuous and stable operation of the system. In edge computing environments, especially image acquisition systems used for wild animal monitoring, there is often a high concentration of image triggering events (such as animal group migration, gathering activities, etc.) within a certain period of time. This will cause a large number of compression and encryption tasks to be triggered simultaneously, causing cache resources to be consumed rapidly in a very short period of time. Once the cache exceeds its carrying capacity, not only will subsequent tasks be unable to execute, but it may also cause the early image data processing to be interrupted or overwritten, resulting in permanent loss of batch image data.

[0093] To avoid these issues, this step first compares the cache resource pressure coefficient (reflecting the system's current cache usage stress) with a preset overvoltage risk threshold to determine in real time whether the system has entered a "cache resource high-risk state." Once the cache resource pressure coefficient exceeds the overvoltage risk threshold, the device is deemed to be on the verge of overvoltage, and the system immediately initiates an adaptive response mechanism. This mechanism uses the current pressure level as an input parameter and prunes image frames (large numbers of images generated by the same event) that are triggered continuously within a short period of time. This pruning process uses evenly spaced frame extraction to ensure that the temporal continuity of the event and the image representativeness are preserved as much as possible while reducing cache load.

[0094] The core value of this approach lies in: preventing data processing crashes caused by cache overflow; and avoiding the severe loss of image information caused by a "blind deletion" strategy. By adjusting the deletion ratio and cache pressure in tandem, the system implements a resource-aware, proactive image processing pruning mechanism. This highly intelligent, practical, and engineering-scalable mechanism is a key guarantee for the stable operation of edge intelligence systems.

[0095] The "edge computing-driven wildlife image compression and encryption transmission method" proposed in this paper effectively addresses the cache conflicts and data failure issues inherent in the prior art, which arise from the serial scheduling of image compression and encryption tasks under resource-constrained conditions. This method comprehensively perceives the dynamic state of cache resource usage by constructing a cache access density curve, a cache response lag indicator, and a load increase rate. This integration generates a normalized cache resource pressure coefficient, enabling quantitative assessment of cache resource stress and potential failure risk. When the system's cache resource pressure coefficient exceeds a risk threshold, it automatically enters a high-risk processing state and, based on the current resource pressure level, implements an adaptive uniform pruning strategy for image frames triggered by high-frequency events. This strategy reduces cache load while preserving event temporal continuity and image integrity. This solution shifts from passive cache overflow response to active cache risk prediction and regulation, significantly improving system stability, image data integrity, and cache resource utilization efficiency for edge devices in high-density image processing scenarios. It holds great promise for engineering applications and widespread adoption.

[0096] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0097] The above description is merely illustrative of certain exemplary embodiments of the present invention. It goes without saying that those skilled in the art will be able to modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and description are illustrative in nature and should not be construed as limiting the scope of protection of the claims.

[0098] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0099] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0100] In the several embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection of some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0101] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0102] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0103] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program code, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0104] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. Edge computing driven wildlife image compression and encryption transmission method, characterized in that: The following steps are involved: S101. Build a cache access density curve based on the task trigger frequency and cache call count of the edge computing device per unit time, record the time series of the maximum available space and the current available space in the cache area, and extract the cache fluctuation period and instantaneous compression amplitude; S102: Calculate the difference between the cache allocation duration and the release delay, construct a cache response hysteresis index, perform weighted fusion on the cache response hysteresis index and the cache access density curve, and generate an intermediate feature vector representing the continuity of cache resource occupancy and the degree of release imbalance; The steps to construct the cache response lag indicator and generate the intermediate feature vector are: In each image compression and encryption process, the time point when the current task applies for cache resources and the time point when the corresponding cache resources are released for the task are recorded respectively, and the time difference between the two is calculated to obtain the cache response lag corresponding to the task; Normalize the cache response lag with the cache access density curve within the time period of the task, and perform a weighted combination based on the preset fusion weight factor. Based on the cache response lag index generated during the continuous processing of multiple tasks, we statistically obtained three dimensional indicators: lag peak, average lag level, and fluctuation amplitude. These indicators serve as basic parameters for measuring cache usage continuity and resource recovery efficiency. The three-dimensional hysteresis statistical features are fused into a multi-dimensional intermediate feature vector to characterize the balance of cache resource usage and the release hysteresis distribution characteristics of edge computing devices under intensive image processing conditions. S103: Collect the intensity change trend of the image trigger event and the average accumulation time of the corresponding processing task, calculate the load increase rate of the cache processing, and introduce a task accumulation saturation function. The task accumulation saturation function is a Sigmoid function. It performs nonlinear amplification mapping on the cache resource pressure under the sudden event and identifies the cache limit approach behavior; S104: Fusing the cache access density curve, the intermediate feature vector, and the load increase rate to construct a multi-dimensional pressure state vector, and mapping the pressure state vector to a cache resource pressure coefficient ranging from 0 to 1 through normalization. S105. When the pressure index exceeds the overpressure risk threshold, the edge device is determined to have entered a high-risk state for cache resources. In the high-risk state for cache resources, based on the current cache resource pressure level of the edge device, a set of image frames of the same continuous event that is continuously triggered within a short period of time is adaptively and uniformly deleted.

2. The edge computing-driven wildlife image compression and encryption transmission method according to claim 1 is characterized in that: The steps of constructing a cache access density curve and extracting cache resource dynamic usage characteristics further include: Based on the number of image trigger requests received per second by the image processing module and the number of compression or encryption processes actually initiated, a corresponding relationship between task trigger frequency and cache call frequency is established. The cache call density per unit time is calculated using a sliding time window to form a preliminary cache access density curve. During cache resource usage, the maximum available space and the current remaining available space in the cache area are continuously monitored to generate time series data. The periodic fluctuation characteristics of cache space changes are extracted through Fourier transform or wavelet analysis. Combining the cache call density change slope and cache space fluctuation gradient, the instantaneous compression amplitude of cache resources per unit time is calculated, and the degree of cache burst shrinkage is determined based on the compression amplitude.

3. The edge computing driven wildlife image compression and encryption transmission method according to claim 1 is characterized in that: The steps to calculate the cache load increase rate and extract the cache limit approaching behavior include: Continuously monitor the intensity trend of image acquisition trigger events, construct an event intensity curve based on the number and frequency of trigger events per unit time, and simultaneously record the average waiting time for each image processing task from the time it enters the processing queue to the time it is scheduled for execution; The cache processing load growth rate is calculated based on the product of the change gradient of the event intensity curve and the growth rate of the average task accumulation time. Construct a task accumulation saturation function that includes nonlinear mapping features and perform weighted amplification on the load increase rate; The processed nonlinear load increase value is used as the input feature and combined with the usage boundary of the cache resources within the time window to identify the extreme approach behavior of the cache resources reaching the full load state.

4. The edge computing-driven wildlife image compression and encryption transmission method according to claim 1 is characterized in that: The steps of constructing a multi-dimensional pressure state vector and generating a cache resource pressure coefficient include: The cache access density curve representing cache access frequency, the intermediate feature vector representing the degree of cache response lag and release imbalance, and the load growth rate reflecting the dynamic growth trend of image processing pressure are feature aligned and time-window synchronized to construct a multidimensional feature dataset at a unified time scale. Perform feature fusion on feature datasets to generate a pressure state vector that comprehensively reflects the degree of cache resource usage. Combined with a preset dynamic weight adjustment mechanism, higher calculation weights are assigned to feature dimensions that are more sensitive to impact. The pressure state vector is normalized and mapped to a cache resource pressure coefficient ranging from 0 to 1, and the cache resource pressure coefficient is used as a risk quantification reference indicator.

5. The edge computing driven wildlife image compression and encryption transmission method according to claim 1 is characterized in that: When cache resources are at high risk, adaptively and evenly delete the image frames of the same continuous event that is triggered continuously within a short period of time. The specific steps are as follows: When the cache resource enters a high-risk state, the image deletion ratio coefficient that needs to be deleted in the image frame set is calculated based on the current cache resource pressure coefficient and the preset overpressure risk threshold. The calculation formula is as follows: ,in: is the cache resource pressure coefficient, indicating the current cache pressure intensity. ; is the overvoltage risk threshold of cache resources. The deletion mechanism is triggered when is the deletion sensitivity adjustment factor, which is used to control the nonlinear response speed of the deletion intensity. ; is the image deletion ratio coefficient, which indicates the ratio that needs to be deleted from the image frame set.

6. The edge computing-driven wildlife image compression and encryption transmission method according to claim 5 is characterized in that: Suppose the set of image frames of the same event that are triggered continuously in a short period of time is ,in N is the original frame number, according to the image deletion ratio coefficient Determine the number of frames that should be retained. The expression for determining the number of frames that should be retained is: ,in: K The number of image frames to be retained is reduced from the original number of frames according to the deletion ratio N Calculation in Indicates rounding up.

7. The edge computing driven wildlife image compression and encryption transmission method according to claim 6 is characterized in that: Keep them at even intervals K Frame images, forming a set of image frames retained after deletion , ,in: ,in: The remaining i Frame image; i Indicates the index of the image frame retained after deletion, .

Images