A multi-view millimeter wave security radar cooperative wide-range target joint detection method for oil fields

Abstract

A kind of oilfield large-scale target joint detection method of multi-view millimeter wave security radar cooperation belongs to millimeter wave security radar and multi-sensor cooperative sensing technical field, in order to solve the problem of time synchronization deficiency, target correlation inaccuracy, fusion precision limitation and trajectory continuity in oilfield large-scale multi-radar target detection, the present application determines the global ID and continuous trajectory information of target by data standardization processing, GPS time stamp time synchronization, target correlation processing, information fusion under fusion weight strategy and global target tracking, and is suitable for oilfield perimeter security, well site equipment monitoring, oil and gas reservoir area personnel and vehicle target detection and other fields.

Description

Technical Field

[0001] This invention belongs to the field of millimeter-wave security radar and multi-sensor collaborative sensing technology, specifically involving the joint detection technology of large-scale targets in oil fields using multi-view millimeter-wave security radar. Background Technology

[0002] As security requirements in oilfield production areas, well sites, and oil and gas storage areas continue to increase, millimeter-wave radar-based target detection technology is increasingly being used for perimeter monitoring of moving targets such as personnel and vehicles. Compared to video surveillance, millimeter-wave security radar is less affected by environmental factors such as lighting, rain, and fog, making it suitable for target perception in large-scale, all-weather scenarios within oilfields. To expand detection coverage, existing technologies are beginning to employ multi-radar collaboration, radar fusion with other sensors, and target trajectory tracking to improve regional monitoring capabilities.

[0003] Among existing solutions, some achieve cross-regional target tracking through data transmission between multiple radars, track fusion, or continuous tracking models; others combine cameras and millimeter-wave radar for temporal registration, spatial registration, and target matching; still others employ multi-sensor fusion coefficients, dynamic weighted fusion, optimal matching, or filtering tracking methods to improve target localization and trajectory stability. These solutions can improve the insufficient detection capabilities of single sensors in general traffic, moving target tracking, or wide-area surveillance scenarios, and provide a certain technical foundation for multi-source target perception.

[0004] However, oilfield sites are characterized by large geographical spans, dispersed radar deployments, simultaneous activities of multiple personnel and vehicles, equipment obstruction, and frequent cross-regional movement. Existing multi-radar detection solutions still suffer from insufficient adaptability. On the one hand, data collected by dispersed radars often suffers from inconsistent timestamps and misaligned data frames, which can easily affect subsequent target association. On the other hand, simple threshold matching or local matching methods are difficult to reliably identify the same target in complex multi-target scenarios, easily leading to false and false matches. Furthermore, existing fusion methods often focus on equal weighting or fixed rule processing, failing to fully consider factors such as detection confidence and the distance between the target and the radar, resulting in limited improvement in the accuracy of target position and velocity after fusion.

[0005] Furthermore, when large-scale targets in oilfields move between different radar coverage areas, existing solutions are prone to trajectory interruptions, target ID switching, or repeated identification, making it difficult to form continuous target trajectories for oilfield scenarios. Therefore, a joint detection method for multi-view millimeter-wave security radar in oilfields is needed. This method improves the coverage, positioning accuracy, and trajectory continuity of joint target detection in large-scale oilfield scenarios by standardizing radar detection data, time synchronization based on GPS timestamps, target association within the same time frame, information fusion under a fusion weight strategy, and global target tracking processing. Summary of the Invention

[0006] To address the problems of insufficient time synchronization, inaccurate target association, limited fusion accuracy, and poor trajectory continuity in existing large-area multi-radar target detection in oilfields, this invention proposes the following solution: A method for joint detection of large-area targets in oil fields using multi-view millimeter-wave security radar, the method comprising: S1. Obtain radar detection data from multi-view millimeter-wave security radar at the oilfield site, and perform data standardization processing on the radar detection data to obtain a radar data list; S2. Based on the GPS timestamps in the radar data list, perform time synchronization processing on the detection data of each radar to obtain synchronized multi-radar time frame data. S3. Based on the synchronized multi-radar time frame data, target association processing is performed on targets detected by different radars within the same time frame to obtain the multi-radar detection association result of the same target. S4. Based on the multi-radar detection association results of the same target, the position and velocity of the same target are fused according to the fusion weight strategy to obtain the fused target detection result. S5. Based on the fused target detection results, perform global target tracking, determine the global ID and continuous trajectory information of the target, and output the joint detection results of targets over a large area of ​​the oilfield.

[0007] Furthermore, the acquisition of radar detection data from multi-view millimeter-wave security radar at the oilfield site as described in S1 includes: collecting raw detection data from the oilfield site using at least two and no more than eight multi-view millimeter-wave security radars. The raw detection data from each radar includes a GPS timestamp, target location, velocity, and detection confidence level.

[0008] Furthermore, the data standardization processing described in S1 includes: removing outliers and standardizing the data format of the radar detection data, and converting the target positions detected by different radars to a unified oilfield scene coordinate system to obtain a radar data list containing GPS timestamps, target positions, velocities, and detection confidence in the unified oilfield scene coordinate system.

[0009] Furthermore, the time synchronization processing described in S2 includes: collecting the timestamps of all networked radar data in the radar data list, deduplicating and sorting them to form a unified time frame sequence; for each unified time frame, matching the detection data frame with the smallest time difference for each radar; when the time difference between the detection data frame and the unified time frame is not greater than the preset time synchronization tolerance, incorporating the detection data frame into the current unified time frame to obtain synchronized multi-radar time frame data.

[0010] Furthermore, the target association processing described in S3 includes: constructing a cost matrix based on the Euclidean distance between each pair of radars for the synchronized single-time-frame multi-radar detection data; using the Hungarian algorithm to perform a global optimal solution on the cost matrix to obtain matching pairs of targets detected by multiple radars; and filtering valid association results through a maximum association distance threshold.

[0011] Furthermore, in the target association processing described in S3, the matrix elements of the cost matrix are the Euclidean distance between the position coordinate vectors of different radar-detected targets. When the Euclidean distance is greater than the maximum association distance threshold, the corresponding matrix element is set to a preset invalid matching value, and the detected target pair is not included in the valid association result.

[0012] Furthermore, S3 describes obtaining the multi-radar detection association results for the same target by: merging the effective association results obtained between two radars, and merging matching pairs with a common target index or target position that meets the maximum association distance threshold into the multi-radar detection association results for the same target.

[0013] Furthermore, the fusion weighting strategy described in S4 includes a confidence-weighted strategy and an inverse-range-weighted strategy. When using the confidence-weighted strategy, the fusion weight is calculated based on the detection confidence of each radar detection result of the same target after association, and the radar detection result with higher detection confidence corresponds to a larger fusion weight. When using the inverse-range-weighted strategy, the fusion weight is calculated based on the distance between the position of the same target in the unified oilfield scene coordinate system and the corresponding radar deployment position, and the radar detection result with smaller distance corresponds to a larger fusion weight.

[0014] Furthermore, the global target tracking described in S5 includes: performing global ID matching on the fused target detection result based on the distance between the fused target position and the predicted position of the tracked target; if the matching fails, the fused target detection result is identified as the first new target, a unique global ID is assigned, and the Kalman filter state and covariance matrix are initialized; if the matching succeeds, the global ID of the tracked target is used, and the state prediction and state update of the Kalman filter are performed, outputting the target's global ID, tracked position, velocity, and continuous trajectory information.

[0015] Based on the same inventive concept, this invention also proposes a computer storage medium storing a computer program, which executes the above-mentioned method for joint detection of large-scale targets in oil fields in coordination with multi-view millimeter-wave security radar when running on a processor.

[0016] Compared with the prior art, the present invention has the following beneficial effects: By acquiring radar detection data from multi-view millimeter-wave security radars at the oilfield site and performing data standardization processing, a unified data foundation is formed for GPS timestamps, target positions, speeds, and detection confidence levels collected by different radars, thereby reducing the impact of data format differences on subsequent joint detection.

[0017] By synchronizing the detection data of each radar based on GPS timestamps, the data from different radars can be aligned to synchronized multi-radar time frames, thereby improving the target association deviation problem caused by misalignment of data frames from dispersed radars. Compared with existing technologies, this process provides a more stable time reference for joint multi-radar target detection.

[0018] By performing target association processing on targets detected by different radars within the same time frame and obtaining multi-radar detection association results for the same target, targets detected by multi-view radars can be merged into corresponding target objects, thereby reducing the risk of mismatch and missed match in complex multi-target scenarios. Compared with existing technologies, this processing can improve the accuracy of target association for personnel and vehicles in oilfields.

[0019] By fusing the position and velocity of the same target according to the fusion weight strategy, the associated multi-radar detection results can be weighted by combining the detection confidence or the distance between the target and the radar deployment location, thereby improving the problem of insufficient positioning and velocity measurement accuracy caused by fixed rules or equal weight fusion.

[0020] By performing global target tracking based on the fused target detection results, the global ID and continuous trajectory information of the target are determined, enabling the target to maintain unified identification and continuous tracking when moving between different radar coverage areas. This reduces trajectory interruptions and target ID switching, and improves the continuity and stability of joint detection results for large-scale targets in the oilfield.

[0021] This invention features multi-radar time synchronization, target association, information fusion, and global target tracking capabilities. It can improve the coverage, positioning accuracy, and trajectory continuity of joint target detection in large-scale oilfield scenarios and is applicable to fields such as oilfield perimeter security, well site equipment monitoring, and personnel and vehicle target detection in oil and gas storage areas. Attached Figure Description

[0022] Figure 1 This is a flowchart of the method for joint detection of large-scale targets in oil fields using multi-view millimeter-wave security radar in collaboration with the implementation method. Figure 2 This is the overall system architecture diagram described in the implementation method; Figure 3 This is a flowchart of the core processing of multi-radar joint detection as described in the implementation method. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] Implementation Method 1 like Figure 1 As shown, a method for joint detection of large-area targets in oil fields using multi-view millimeter-wave security radar is described, the method comprising: S1. Obtain radar detection data from multi-view millimeter-wave security radar at the oilfield site, and perform data standardization processing on the radar detection data to obtain a radar data list; S2. Based on the GPS timestamps in the radar data list, perform time synchronization processing on the detection data of each radar to obtain synchronized multi-radar time frame data. S3. Based on the synchronized multi-radar time frame data, target association processing is performed on targets detected by different radars within the same time frame to obtain the multi-radar detection association result of the same target. S4. Based on the multi-radar detection association results of the same target, the position and velocity of the same target are fused according to the fusion weight strategy to obtain the fused target detection result. S5. Based on the fused target detection results, perform global target tracking, determine the global ID and continuous trajectory information of the target, and output the joint detection results of targets over a large area of ​​the oilfield.

[0025] By sequentially performing data standardization, time synchronization, target association, information fusion, and global target tracking on radar detection data, a continuous processing link for large-scale target detection in oil fields can be formed, thereby obtaining joint detection results with global ID and continuous trajectory information.

[0026] Furthermore, the acquisition of radar detection data from multi-view millimeter-wave security radar at the oilfield site as described in S1 includes: collecting raw detection data from the oilfield site using at least two and no more than eight multi-view millimeter-wave security radars. The raw detection data from each radar includes a GPS timestamp, target location, velocity, and detection confidence level.

[0027] By collecting raw detection data including GPS timestamps, target locations, speeds, and detection confidence levels from at least two and no more than eight multi-view millimeter-wave security radars, a complete data source can be provided for subsequent time synchronization, target association, and information fusion.

[0028] Furthermore, the data standardization processing described in S1 includes: removing outliers and standardizing the data format of the radar detection data, and converting the target positions detected by different radars to a unified oilfield scene coordinate system to obtain a radar data list containing GPS timestamps, target positions, velocities, and detection confidence in the unified oilfield scene coordinate system.

[0029] Preferably, the GPS timestamps, target positions, speeds, and detection confidence levels recorded for the same target in the radar data list maintain a corresponding relationship.

[0030] By removing outliers, standardizing data formats, and transforming the coordinate system of a unified oilfield scene, different radar detection data can have a unified data representation basis, which facilitates the subsequent association and fusion processing of targets within the same time frame.

[0031] Furthermore, the time synchronization processing described in S2 includes: collecting the timestamps of all networked radar data in the radar data list, deduplicating and sorting them to form a unified time frame sequence; for each unified time frame, matching the detection data frame with the smallest time difference for each radar; when the time difference between the detection data frame and the unified time frame is not greater than the preset time synchronization tolerance, incorporating the detection data frame into the current unified time frame to obtain synchronized multi-radar time frame data.

[0032] By collecting, deduplicating, and sorting the timestamps of data from each network of radars, and filtering the detection data frames corresponding to each radar according to the time synchronization tolerance, the detection data of different radars can be aligned to a unified time frame, thereby providing a time basis for target association processing within the same time frame.

[0033] Furthermore, the target association processing described in S3 includes: constructing a cost matrix based on the Euclidean distance between each pair of radars for the synchronized single-time-frame multi-radar detection data; using the Hungarian algorithm to perform a global optimal solution on the cost matrix to obtain matching pairs of targets detected by multiple radars; and filtering valid association results through a maximum association distance threshold.

[0034] By constructing a cost matrix based on the Euclidean distance of the target location and using the Hungarian algorithm for global optimal solution, the matching relationship between different radar-detected targets can be determined within a single time frame, thus obtaining effective association results for subsequent merging and fusion.

[0035] Furthermore, in the target association processing described in S3, the matrix elements of the cost matrix are the Euclidean distance between the position coordinate vectors of different radar-detected targets. When the Euclidean distance is greater than the maximum association distance threshold, the corresponding matrix element is set to a preset invalid matching value, and the detected target pair is not included in the valid association result.

[0036] By using the Euclidean distance between the position coordinate vectors of targets detected by different radars as the matrix elements of the cost matrix, and by using the maximum association distance threshold to exclude invalid matches, unreasonable target matching relationships can be restricted, thereby improving the reliability of effective association results.

[0037] Furthermore, S3 describes obtaining the multi-radar detection association results for the same target by: merging the effective association results obtained between two radars, and merging matching pairs with a common target index or target position that meets the maximum association distance threshold into the multi-radar detection association results for the same target.

[0038] By merging the effective correlation results between pairs of radars, the detection results of multiple radars belonging to the same target can be organized into a unified multi-radar detection correlation result, thereby providing a data foundation for the fusion of the position and velocity of the same target.

[0039] Furthermore, the fusion weighting strategy described in S4 includes a confidence-weighted strategy and an inverse-range-weighted strategy. When using the confidence-weighted strategy, the fusion weight is calculated based on the detection confidence of each radar detection result of the same target after association, and the radar detection result with higher detection confidence corresponds to a larger fusion weight. When using the inverse-range-weighted strategy, the fusion weight is calculated based on the distance between the position of the same target in the unified oilfield scene coordinate system and the corresponding radar deployment position, and the radar detection result with smaller distance corresponds to a larger fusion weight.

[0040] By employing a confidence-weighted strategy or an inverse range-weighted strategy to calculate the fusion weights, the detection results of multiple radars for the same target can be weighted and fused based on the detection confidence or the distance between the target and the radar deployment location, thereby obtaining the fused target detection result.

[0041] Furthermore, the global target tracking described in S5 includes: performing global ID matching on the fused target detection result based on the distance between the fused target position and the predicted position of the tracked target; if the matching fails, the fused target detection result is identified as the first new target, a unique global ID is assigned, and the Kalman filter state and covariance matrix are initialized; if the matching succeeds, the global ID of the tracked target is used, and the state prediction and state update of the Kalman filter are performed, outputting the target's global ID, tracked position, velocity, and continuous trajectory information.

[0042] By performing global ID matching based on the distance between the fused target position and the predicted position of the tracked target, and combining Kalman filter state prediction and state update, the target can be uniformly identified and continuously tracked, thereby outputting the target's global ID, tracked position, velocity, and continuous trajectory information.

[0043] The method described in this embodiment can be executed by a processor calling a computer program, which can be stored in a computer storage medium. When the computer program is executed by the processor, the aforementioned method for joint detection of large-scale targets in oil fields using multi-view millimeter-wave security radar can be realized.

[0044] Implementation Method 2 This embodiment provides a method for joint detection of large-scale targets in oilfields using multi-view millimeter-wave security radar. This method is based on the construction of a fully automated closed loop that integrates spatiotemporal synchronization of multi-radar data, target association within the same frame, multi-source information fusion, global trajectory tracking, parameter self-optimization, and visualization output. It covers processing levels such as parameter configuration, data synchronization, target association, information fusion, global tracking, and visualization monitoring. Data exchange, result feedback, and reverse optimization are performed between each level to achieve joint processing of multi-view millimeter-wave security radar detection data in large-scale oilfield scenarios. Figure 2 The overall system architecture corresponding to the method described in this embodiment is shown.

[0045] When the method of this embodiment is executed, it is based on multi-radar spatiotemporal synchronization, with global target association as the core, adaptive information fusion as the position and velocity fusion method, Kalman filter global tracking as the continuous trajectory maintenance method, and parameter self-optimization as the iterative processing method. The parameter configuration process allows for flexible setting and updating of core parameters such as time synchronization tolerance, maximum correlation distance, fusion weight strategy, Kalman filter parameters, and the number of networked radars, outputting personalized configuration parameters adapted to oilfield scenarios. The data synchronization process enables the acquisition, preprocessing, and high-precision time synchronization of raw data from multiple radars, aligning the time frames of multi-radar data to provide a unified spatiotemporal reference for subsequent processing. The target association process performs global optimal matching and effective association filtering of targets from synchronized multi-radar detection data of the same frame, outputting multi-radar detection association results for the same target. The information fusion process uses an adaptive weighting strategy to fuse position and velocity information from the associated multi-radar target detection data, outputting fused detection results. The global tracking process uses Kalman filtering to perform unified global ID management of the fused targets, predicting and updating target states, and outputting continuous target trajectories and tracking results. The visualization monitoring process integrates real-time display of radar deployment status, raw detection results, fused target information, and tracking trajectories, as well as interactive functions such as operation log storage, online parameter configuration, and trajectory playback.

[0046] In the spatiotemporal synchronization of multi-radar data, to address the issues of clock asynchrony and data spatiotemporal misalignment among multiple radars deployed over a large area in oilfields, a GPS timestamp-based time synchronization mechanism is constructed. A time synchronization tolerance threshold is set, and nearest neighbor time matching is used to achieve unified time frame alignment of multi-radar data, realizing high-precision spatiotemporal synchronization of multi-view radar data and providing a unified time reference for subsequent target association. This time synchronization process achieves millisecond-level time synchronization accuracy through nearest neighbor time matching and tolerance threshold control, with a synchronization success rate of ≥99%, making it suitable for large-scale, distributed multi-radar network scenarios in oilfields.

[0047] During the spatiotemporal synchronization of multi-radar data, the original data timestamps of all networked radars are first collected, deduplicated, and sorted to form a unified time frame sequence for the system; then, for each unified time frame... Calculate timestamps for all data frames of a single radar. The absolute difference is used to match the detection frame with the smallest time difference. The valid matching judgment rule is:

[0048] in, For the timestamp of a single radar data frame, The preset time synchronization tolerance is 0.1 seconds by default. If the judgment rule is met, the detection data of the radar is included in the current time frame; otherwise, the radar has no valid detection data in the current time frame. Finally, the synchronized multi-radar time frame dataset is output.

[0049] In the multi-view radar target association process, to address the matching problem of multiple radar-detected targets within the same time frame, a cost matrix based on the Euclidean distance of the target positions is constructed. The globally optimal matching pair is solved using the Hungarian algorithm, and valid association results are filtered by combining the maximum association distance threshold, thus achieving multi-view radar matching of the same target. This target association process replaces the simple threshold matching method, using the Hungarian algorithm to achieve globally optimal matching. In complex scenarios with multiple personnel and vehicles in oil fields, the target association accuracy is improved by ≥30%, and the false matching rate is reduced by more than 80%.

[0050] In the multi-view radar target association process, for M detected targets of radar 1 and N detected targets of radar 2 within the same time frame, an M×N cost matrix C is constructed, where the matrix elements are calculated using the following formula:

[0051] in, For radar 1 The position coordinate vector of each target For radar 2 The position coordinate vectors of each target; if (Maximum associated distance threshold, default 50 meters), then Set as The maximum value of the cost matrix is ​​used to exclude invalid matches. The Hungarian algorithm is then applied to the cost matrix. A global optimization solution is performed to obtain the optimal matching pair of row indexes and column indexes. For matching pairs Only when If the association is deemed valid, it will be included in the subsequent information fusion processing flow.

[0052] In the adaptive fusion of multi-radar target information, for the same target detection data from multiple radars after association, two switchable adaptive weighted fusion strategies are constructed: confidence-weighted and inverse-range-weighted. The fusion weights of the detection results from each radar are dynamically calculated, and the target's position and velocity information are weighted and averaged while retaining the highest detection confidence level, thus achieving the fusion processing of multi-radar detection information. This information fusion process replaces the equal-weighted fusion method and can switch fusion strategies according to the radar deployment scenario and detection environment. The fused target positioning error is reduced by ≥40%, and the velocity measurement accuracy is improved by ≥35%.

[0053] In the adaptive fusion of multi-radar target information, for the correlated... For each radar detection result, calculate the weight of each detection result. It supports two strategies: The confidence-weighted strategy is as follows:

[0054] in, For the first The confidence level of each radar detection result.

[0055]

[0056] in, For the first The radar detects the distance between the target and the radar itself. Minimum value ( ), used to avoid a denominator of 0.

[0057] The target location after fusion is:

[0058] The target speed after fusion is:

[0059] in, , The first The target position and velocity vectors detected by each radar are used. The confidence score after fusion is the highest confidence score among the multiple radar detection results, i.e.:

[0060] In the global target tracking process based on Kalman filtering, a Kalman filter tracker based on a constant-velocity model is constructed to meet the needs of large-scale oilfield coverage. This tracker performs unified global ID management on the fused target detection results, completes target state prediction and updating, and achieves continuous trajectory tracking of targets within the coverage area of ​​multiple radars. This global target tracking process is driven by multi-radar fusion results, achieving globally unique target ID allocation and continuous tracking across the entire area. The trajectory continuity rate is improved to ≥95%, and the target loss rate across radar regions is reduced to below 1%.

[0061] In global target tracking based on Kalman filtering, the target state vector is defined as:

[0062] in, , The target plane position coordinates, , For the goal , The velocity component in the direction. In the core matrix definition, the time interval is... State transition matrix Using a constant-velocity motion model, it can be represented as:

[0063] Observation matrix (Mapping the state to a location observation):

[0064] Process noise covariance matrix Observation noise covariance matrix It can be flexibly adjusted and optimized through the parameter configuration layer.

[0065] The state prediction process is as follows:

[0066]

[0067] in, , Given the target state and covariance matrix at the previous time step, , These are the predicted state and covariance values ​​at the current moment.

[0068] During the state update process, the Kalman gain is calculated as follows:

[0069] Status updated to:

[0070] Covariance updated to:

[0071] in, The fused target location observations It is an identity matrix.

[0072] During the global ID management process, a unique global ID is assigned to a new target that appears for the first time, and the Kalman filter state and covariance matrix are initialized; for targets with existing IDs, the prediction and update process is executed after matching is completed, and the complete trajectory of the target is continuously maintained.

[0073] This implementation method supports network access for up to eight multi-view millimeter-wave security radars during the execution of the multi-radar joint detection main control process. It is compatible with both synthetic simulation data and real radar data from the oilfield, completing the entire automated process from data synchronization, target association, information fusion to global tracking. It also includes a built-in parameter self-optimization training process to achieve unified output of joint detection and tracking results. This main control process supports collaborative network processing of up to eight radars, expanding the effective coverage of a single radar by more than four times. The entire process has a processing latency of ≤100ms, meeting the real-time requirements of oilfield industrial scenarios, while also being compatible with dual data source input.

[0074] During multi-data source compatibility processing, it supports seamless switching between radar simulation synthetic data and real radar data from the oilfield. It has a built-in standardized data adaptation interface, allowing data access without additional development. In the fully automated scheduling process, it sequentially executes data synchronization, target association, information fusion, and global tracking according to time frames, without any manual intervention, achieving automated joint detection. During trainable optimization, it includes a built-in gradient descent-based parameter optimization module, which can iteratively train and optimize core parameters such as Kalman filter noise parameters and fusion weight strategies based on labeled real trajectory data.

[0075] The multi-view millimeter-wave security radar collaborative oilfield large-scale target joint detection method of this embodiment is specifically implemented in steps S1-S8, with each step executed sequentially, data exchanged, and reverse optimized to form a complete joint detection closed loop. Figure 3 The core processing flow of multi-radar joint detection described in this embodiment is shown.

[0076] S1: System Initialization and Parameter Configuration. Start the multi-radar joint detection system, complete hardware adaptation and software module initialization, supporting automatic adaptation to CPU / GPU accelerated environments; configure core system parameters in the visual interface, including time synchronization tolerance, maximum correlation distance threshold, fusion weight strategy, Kalman filter time interval and noise parameters, and the number of networked radars (maximum 8 radars); initialize each core functional module, complete communication connections and data interface adaptation for the multi-radar network, preparing for subsequent processing.

[0077] S2: Raw Data Acquisition and Preprocessing from Multiple Radars. This involves simultaneously acquiring raw detection data from multi-view millimeter-wave security radars at the oilfield site. The raw data for each radar includes core fields such as GPS timestamp, target location, velocity, and detection confidence level. The raw data is preprocessed to remove outliers, standardize data formats, and output a radar data list adapted to the system's processing format.

[0078] S3: High-precision time synchronization of multi-radar data. Collect timestamps from all networked radar data, deduplicate them, and sort them to form a unified time frame sequence. For each unified time frame, match the detection data frame with the smallest time difference for each radar. Filter valid data using a time synchronization tolerance threshold to complete the time frame alignment of multi-radar data. Output synchronized multi-radar time frame data, where each time frame contains a unified timestamp and valid detection results from each radar.

[0079] S4: Multi-radar target association matching in the same time frame. For the synchronized single-time frame multi-radar detection data, a cost matrix based on the Euclidean distance between target positions is constructed between each pair of radars; the Hungarian algorithm is used to solve the cost matrix globally to obtain matching pairs of multi-radar detected targets; valid association results are filtered by the maximum association distance threshold, invalid matching pairs are excluded, and the multi-radar detection association results for the same target are output.

[0080] S5: Adaptive Weighted Fusion of Multi-Radar Target Information. For the multi-radar detection results of the same target after association, the fusion weight of each radar detection result is calculated according to a preset weighting strategy, which is a confidence-weighted strategy or an inverse range-weighted strategy; the target's position and velocity information are weighted and averaged based on the fusion weight to obtain the fused target position and velocity; the highest confidence among the multi-radar detection results is taken as the confidence of the fused target, a temporary identifier is assigned to each fused target, and the fused target detection result is output.

[0081] S6: Global Target Tracking Based on Kalman Filter. For the fused target detection results, global ID matching is performed: if it is a new target appearing for the first time, a unique global ID is assigned, and the Kalman filter state and covariance matrix are initialized; if it is an existing target, the corresponding global ID is matched; for the tracked target, the Kalman filter state prediction step is executed to obtain the predicted target position; using the fused target position as the observation value, the Kalman filter state update step is executed to correct the target state and covariance matrix; the target's global ID, tracked position, velocity, and complete trajectory information are output.

[0082] S7: System Parameter Self-Optimization Training. Collect radar detection datasets of oilfield scenes with real trajectory annotations to construct a training dataset; use the mean square error between the fused target position and the real trajectory position as the loss function, and employ the Adam optimizer to iteratively train and optimize learnable parameters such as the noise parameters of the Kalman filter and the fusion weight strategy; after training is completed, save the optimized system parameters and update them to the corresponding modules.

[0083] S8: Visualization and Data Storage of Detection Results. The visualization interface displays in real-time the deployment locations of the networked radars, the original detection results of each radar, the fused target detection results, the global target tracking trajectory, and the motion status. It automatically records all system operation data, including timestamps, original radar data, correlation results, fusion results, and tracking trajectories, forming a standardized operation log that supports local storage and historical data tracing. It also supports interactive functions such as target intrusion alarms, trajectory playback, and online parameter modification, adapting to the actual needs of oilfield security operations and maintenance.

[0084] The above detailed description of the technical solution provided by the present invention is intended to highlight the advantages and benefits of the technical solution provided by the present invention. However, the above detailed embodiments are not intended to limit the scope of protection of the present invention. Any reasonable modifications and improvements to the present invention, recombination of embodiments, and equivalent substitutions based on the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0085] Those skilled in the art will understand that the above description is merely a preferred embodiment of the present invention, and the features described in the various embodiments and / or claims disclosed in the present invention can be combined or combined in various ways, even if such combinations or combinations are not explicitly described in the disclosure of the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle scope of the present invention should be considered to fall within the protection scope of the present invention.

[0086] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its spirit and scope. Thus, if these modifications and modifications of the invention fall within the scope of the claims and their equivalents, the invention is also intended to include these modifications and modifications.

Claims

1. A method for joint detection of large-scale targets in oil fields using multi-view millimeter-wave security radar, characterized in that, The method includes: S1. Obtain radar detection data from multi-view millimeter-wave security radar at the oilfield site, and perform data standardization processing on the radar detection data to obtain a radar data list; S2. Based on the GPS timestamps in the radar data list, perform time synchronization processing on the detection data of each radar to obtain synchronized multi-radar time frame data. S3. Based on the synchronized multi-radar time frame data, target association processing is performed on targets detected by different radars within the same time frame to obtain the multi-radar detection association result of the same target. S4. Based on the multi-radar detection association results of the same target, the position and velocity of the same target are fused according to the fusion weight strategy to obtain the fused target detection result. S5. Based on the fused target detection results, perform global target tracking, determine the global ID and continuous trajectory information of the target, and output the joint detection results of targets over a large area of ​​the oilfield.

2. The method according to claim 1, characterized in that, The acquisition of radar detection data from multi-view millimeter-wave security radar at the oilfield site, as described in S1, includes: collecting raw detection data from the oilfield site using at least two and no more than eight multi-view millimeter-wave security radars. The raw detection data from each radar includes a GPS timestamp, target location, velocity, and detection confidence level.

3. The method according to claim 1, characterized in that, The data standardization process described in S1 includes: removing outliers and standardizing the data format of the radar detection data, and converting the target positions detected by different radars to a unified oilfield scene coordinate system to obtain a radar data list containing GPS timestamps, target positions, velocities, and detection confidence in the unified oilfield scene coordinate system.

4. The method according to claim 1, characterized in that, The time synchronization process described in S2 includes: collecting the timestamps of all networked radar data in the radar data list, deduplicating and sorting them to form a unified time frame sequence; for each unified time frame, matching the detection data frame with the smallest time difference for each radar; when the time difference between the detection data frame and the unified time frame is not greater than the preset time synchronization tolerance, incorporating the detection data frame into the current unified time frame to obtain synchronized multi-radar time frame data.

5. The method according to claim 1, characterized in that, The target association processing described in S3 includes: constructing a cost matrix based on the Euclidean distance between each pair of radars for the synchronized single-time-frame multi-radar detection data; using the Hungarian algorithm to perform a global optimal solution on the cost matrix to obtain matching pairs of targets detected by multiple radars; and filtering valid association results through a maximum association distance threshold.

6. The method according to claim 5, characterized in that, The cost matrix elements are the Euclidean distances between the position coordinate vectors of different radar-detected targets. When the Euclidean distance is greater than the maximum association distance threshold, the corresponding matrix element is set to a preset invalid matching value, and the detected target pair is not included in the valid association results.

7. The method according to claim 1, characterized in that, S3 describes obtaining the multi-radar detection association results for the same target, which includes: merging the effective association results obtained between two pairs of radars, and merging matching pairs with a common target index or target position that meets the maximum association distance threshold into the multi-radar detection association results for the same target.

8. The method according to claim 1, characterized in that, The fusion weighting strategy described in S4 includes a confidence-weighted strategy and an inverse-range-weighted strategy. When using the confidence-weighted strategy, the fusion weight is calculated based on the detection confidence of each radar detection result for the same target after association, and the radar detection result with higher detection confidence corresponds to a larger fusion weight. When using the inverse-range-weighted strategy, the fusion weight is calculated based on the distance between the position of the same target in the unified oilfield scene coordinate system and the corresponding radar deployment position, and the radar detection result with smaller distance corresponds to a larger fusion weight.

9. The method according to claim 1, characterized in that, S5 describes global target tracking as follows: based on the distance between the fused target position and the predicted position of the tracked target, perform global ID matching on the fused target detection result; if the matching fails, determine the fused target detection result as the first new target, assign a unique global ID, and initialize the Kalman filter state and covariance matrix; if the matching succeeds, use the global ID of the tracked target, perform Kalman filter state prediction and state update, and output the target's global ID, tracked position, velocity, and continuous trajectory information.

10. A computer storage medium having a computer program stored thereon, characterized in that, The computer program, when running on a processor, performs the method according to any one of claims 1 to 9.

Images