Disclosure of Invention
In order to solve the technical problems, the application provides a rocket recovery sub-stage air recognition and tracking method, equipment and a storage medium, which can effectively and accurately recognize and classify the rocket recovery sub-stage, improve the accuracy and reliability of target recognition and effectively track the rocket recovery sub-stage.
A rocket recovery sub-level air recognition and tracking method comprises the following steps:
acquiring data related to rocket recovery sub-level air identification and tracking, and preprocessing the acquired data;
fusing the preprocessed information to extract multi-mode characteristics of rocket recovery sub-stages;
carrying out rocket recovery sub-level tracking and data association by combining a KNN algorithm, comparing and matching target characteristics of a current frame with target characteristics of a previous frame, establishing a rocket recovery sub-level track, and updating a rocket recovery sub-level state;
based on target track and historical motion information of the rocket recovery sub-stage, predicting, locking and tracking rocket recovery sub-stage motion by utilizing a SORT algorithm;
the position, speed and attitude information of the rocket recovery sub-stage are updated in real time, and accurate feedback and decision basis are provided for rocket recovery sub-stage air identification and tracking.
Preferably, the data related to rocket recovery sub-level air identification and tracking comprises: image data, video data, distance data, and speed data.
Preferably, preprocessing the data includes denoising, filtering, and data correction.
Preferably, the multi-modal features in the multi-modal features of the extracted rocket recovery sub-stage include shape, color, texture, and velocity and acceleration of the rocket recovery sub-stage.
Preferably, in the fusing of the preprocessed information, the formula for performing information fusion calculation is as follows:
;
in the method, in the process of the application,is a fusion feature vector; />Image feature weight when features are fused; />Video feature weights at feature fusion; />Distance feature weight when features are fused; />The speed characteristic weight is used for characteristic fusion;a vector representation of image features; />A vector representation that is a feature of the video; />A vector representation that is a distance feature; />Is a vector representation of the velocity features.
Preferably, the performing rocket recovery sub-level tracking and data association by combining with the KNN algorithm, comparing and matching the target feature of the current frame with the target feature of the previous frame, establishing a trajectory of the rocket recovery sub-level, and updating the rocket recovery sub-level state includes:
rocket recovery sub-level detection and recognition are carried out based on the comprehensive feature vectors;
target positioning is carried out based on comprehensive feature vectors, and rocket recovery sub-level positions are obtained
Performing target speed estimation based on the speed feature vector to obtain a rocket recovery sub-level speed estimation value;
performing target tracking based on the rocket recovery sub-level position and the velocity estimation value, and determining the rocket recovery sub-level state;
and carrying out rocket recovery sub-level state comprehensive evaluation based on the fusion feature vector and the target detection result.
Preferably, the performing rocket recovery sub-level state comprehensive evaluation based on the fusion feature vector and the target detection result includes:
weighting calculation is carried out on the fusion feature vector;
normalizing the feature vector;
weighting calculation is carried out on the target detection result:
and (5) comprehensively evaluating and calculating to determine the final recognition and positioning results of the arrow recovery sub-level after fusion.
Preferably, the comprehensive evaluation calculation formula is:
;
in the method, in the process of the application,for final fused target recognition and localization results, < > and so on>For accuracy, ->For confidence level->For the degree of risk->For accuracy weight, ++>Is confidence ofDegree weight (weight of->Is the risk degree weight;
if the accuracy is not less than the accuracy threshold, the accuracy evaluation result is 1, otherwise, the accuracy evaluation result is 0;
if the confidence coefficient is not smaller than the confidence coefficient threshold value, the confidence coefficient assessment result is 1, otherwise, the confidence coefficient assessment result is 0;
if the risk level is not greater than the risk level threshold, the risk level evaluation result is 1, otherwise, the risk level evaluation result is 0.
According to another aspect of the present application, there is also provided a computing device including: a processor, a memory storing a computer program which, when executed by the processor, performs the rocket recovery sub-level air identification and tracking method of any one of claims 1 to 8.
According to another aspect of the application there is also provided a computer readable storage medium having stored thereon computer instructions which, when run on a computer, cause the computer to perform the rocket recovery sub-level air identification and tracking method according to any one of claims 1 to 8.
Compared with the prior art, the application has at least the following beneficial effects:
1. the rocket recovery sub-stage identification method and device can effectively accurately identify and classify rocket recovery sub-stages, and improve the accuracy and reliability of target identification.
2. The application can improve the stability and accuracy of rocket recovery sub-level tracking and ensure target tracking in the rocket sub-level recovery process in a high-speed dynamic environment.
3. The application can also improve the robustness of the rocket sub-level recovery air recognition and tracking system under the complex scene and target shielding condition.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the technical solutions of the present application will be clearly and completely described below with reference to specific embodiments of the present application and corresponding drawings. It will be apparent that the described embodiments are only some, but not all, embodiments of the application. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
As shown in fig. 1, a rocket recovery sub-level air recognition and tracking method comprises the following steps:
and S1, acquiring data related to rocket recovery sub-level air recognition and tracking, and preprocessing the acquired data.
Specifically, data of devices such as a sensor, a camera, a radar and the like are collected, and image data, video data, distance data and speed data are preprocessed. The preprocessing comprises denoising, filtering and data correction, so that the data quality and accuracy are improved.
And S2, fusing the preprocessed information, and extracting multi-mode characteristics of the rocket recovery sub-level.
And the information from different devices such as a sensor, a camera and a radar after preprocessing is fused by utilizing a multi-mode data fusion technology. The extraction of the multi-modal features of the rocket recovery sub-stage includes the shape, color, texture, and velocity and acceleration of the rocket recovery sub-stage.
By comprehensively utilizing multi-mode data of images, videos, distances and speeds, the accuracy and the robustness of target detection and identification can be improved, and rocket sub-level targets can be accurately positioned and classified.
And S3, carrying out rocket recovery sub-level tracking and data association by combining a KNN algorithm, comparing and matching the target characteristics of the current frame with the target characteristics of the previous frame, establishing a rocket recovery sub-level track, and updating the rocket recovery sub-level state.
Specifically, the method comprises the following steps:
step S31, rocket recovery sub-level detection and recognition are carried out based on the comprehensive feature vectors;
s32, performing target positioning based on the comprehensive feature vector to obtain rocket recovery sub-level positions
S33, estimating a target speed based on the speed feature vector to obtain a rocket recovery sub-level speed estimated value;
step S34, performing target tracking based on the rocket recovery sub-level position and the velocity estimation value, and determining the rocket recovery sub-level state;
and S35, comprehensively evaluating the state of the rocket recovery sub-stage based on the fusion feature vector and the target detection result.
Specifically, step S35 includes:
step S351, weighting calculation is carried out on the fusion feature vector;
step S352, normalizing the feature vector;
step S353, performing weighted calculation on the target detection result:
and step 354, comprehensively evaluating and calculating to determine the final recognition and positioning result of the recovered sub-level of the arrow after fusion.
And S4, based on target track and historical motion information of the rocket recovery sub-stage, predicting, locking and tracking rocket recovery sub-stage motion by utilizing an SORT algorithm.
And S5, updating the position, speed and attitude information of the rocket recovery sub-stage in real time, and providing accurate feedback and decision basis for the aerial identification and tracking of the rocket recovery sub-stage.
As one embodiment of the application, the rocket recovery sub-stage is taken as a calculation target, and the implementation process of the rocket recovery sub-stage air recognition and tracking method is as follows:
first, the following parameters are defined:
image data (I): representing image data acquired by the camera.
Video data (V): representing continuous video data acquired by the camera.
Distance data (D): distance data representing the distance between the target and the rocket sub-level acquired by a sensor such as radar.
Speed data (S): data representing the relative velocity of the target and rocket sublevel acquired by a sensor such as radar.
Time (t): and (5) representing the time point of data acquisition under the rocket sublevel recovery time coordinate system.
Target position (P): representing the position of the target in the rocket-level coordinate system.
Target speed [ ]): representing the velocity of the target in the rocket-level coordinate system.
Image characteristics [ ]): representing feature vectors extracted from the image data.
Video feature [ ]): representing feature vectors extracted from video data.
Distance characteristic [ ]): representing feature vectors extracted from the distance data.
Speed characteristics [ ]): representing the feature vector extracted from the velocity data.
Appearance characteristics of%): representing the topographical features of the object, such as aspect ratio, geometry, etc.
Color characterization): representing the color histogram of the object.
Texture characteristics [ ]): representing texture characteristics of the object, such as texture frequency, contrast, etc.
Rocket sub-grade tail flame envelope characteristics): representing the shape and envelope characteristics of the rocket sub-level tail flame.
Target state (X): the state vector representing the object includes information such as position, velocity, etc.
Weight (W): and the weights are used for representing the characteristics of different data sources and are used for fusing information of different data.
Color threshold [ ]): representing threshold parameters for the color feature.
Texture threshold [ ]): representing threshold parameters for texture features.
Envelope threshold [ ]): a threshold parameter representing a characteristic for a rocket-sub-level tail envelope.
Fusion results (R): and representing the target recognition and positioning result after final fusion.
Data preprocessing:
a. preprocessing image data: = PreprocessImage(I);
b. video data preprocessing: = PreprocessVideo(V);
c. distance data preprocessing:= PreprocessDistance(D);
d. preprocessing speed data: = PreprocessSpeed(S)。
wherein PreprocessImage (I) is a function of preprocessing the image data I.
PreprocessVideo (V) is a function of preprocessing the video data V.
PreprocessDistance (D) is a function of preprocessing the distance data D.
PreprocessSpeed (S) a function of preprocessing the speed data S.
And (3) information fusion:
the formula for information fusion calculation is as follows:
;
in the method, in the process of the application,is a fusion feature vector; />Image feature weight when features are fused; />Video feature weights at feature fusion; />Distance feature weight when features are fused; />The speed characteristic weight is used for characteristic fusion;a vector representation of image features; />A vector representation that is a feature of the video; />A vector representation that is a distance feature; />Is a vector representation of the velocity features.
The target detection and identification formula is as follows:
DetectionResult = ObjectDetection(F);
wherein ObjectDetection (F) is a function for performing object detection based on the feature vector F, and DetectionResult is a returned detection result.
Target positioning and tracking includes:
a. target positioning: p= ObjectLocalization (F);
b. target speed estimation: = EstimateVelocity(/>);
c. Target tracking: x=objecttracking (P,);
the ObjectLocalization (F) is a function for performing target positioning based on the feature vector F, and the target position P can be returned by calculation.
EstimateVelocity() Is based on speed characteristics->Function for estimating target speed, by calculating the speed estimate value that can be returned +.>。
ObjectTracking(P, ) For estimating value +.>The function of target tracking is performed, and the target state X can be returned by calculation.
The comprehensive evaluation process is as follows:
R = ObjectEvaluation(F, DetectionResult);
the objection evaluation (F, detectionResult) is a function for performing comprehensive evaluation based on the fusion feature vector F and the target detection result DetectionResult. The function comprehensively considers the weight and the threshold value of each feature, and calculates the evaluation result R of the target according to the set evaluation index. The evaluation result R can help the system to track the target in the rocket sublevel recovery process more accurately.
Specifically, the comprehensive evaluation includes the following procedures:
defining an evaluation index:
accuracy: a, A is as follows;
confidence level: c, performing operation;
degree of risk: d, a step of performing the process;
characteristic weight:,/> ,/> , ..., />;
evaluation index threshold:,/> , />;
the weighted feature vector calculation formula is:
;
the normalized eigenvector calculation formula is:
= Normalize(/>) ;
the target detection result is weighted.
The weighted target detection result is as follows:
;
the comprehensive evaluation calculation formula is as follows:
;
in the method, in the process of the application,for final fused target recognition and localization results, < > and so on>For accuracy, ->For confidence level->For the degree of risk->For accuracy weight, ++>For confidence weight, ++>Is a risk degree weight.
If, during the calculationThen->Otherwise->;
If it isThen->Otherwise->;
If it isThen->Otherwise->。
Wherein, the liquid crystal display device comprises a liquid crystal display device,for the accuracy threshold, ++>For confidence threshold, ++>Is a risk level threshold.
In addition, the method also comprises the step of updating the position, speed and attitude information of the rocket recovery sub-stage in real time:
;
in the method, in the process of the application,for updated target recognition and localization results, < >>For accuracy calculation result +.>For confidence calculation result, ++>The results are calculated for the risk level.
Through the calculation, the comprehensive evaluation result R is obtained and used for tracking and locking rocket sublevel recovery pictures in real time. The dynamic tracking of rocket sublevel can be realized by continuously updating the position and state of the target, and the accuracy, the confidence and the risk degree of the target are evaluated according to the comprehensive evaluation result. This helps to improve the dynamic lock look and feel of the rocket sublevel recovery process.
In the rocket-stage recovery air recognition and tracking method, a KNN algorithm is used for recognizing a rocket-stage target. By using the training data set, the KNN algorithm can calculate the similarity with a known sample according to rocket sub-level data acquired through a wireless sensing network in the rocket sub-level landing area calculated in advance, so as to determine the category of the target. The rocket sub-level can be rapidly and accurately distinguished from other targets, and a starting point is provided for subsequent target tracking.
In the rocket sub-level recovery air recognition and tracking method, the SORT algorithm is used for carrying out real-time tracking on the recognized rocket sub-level targets. By combining a Kalman filtering technology, the SORT algorithm can predict the motion trail and position of the target, and meanwhile, the currently observed target is associated with the existing tracking result by an association filtering technology, so that the tracking accuracy and stability are ensured. The SORT algorithm can reliably track rocket sublevel targets in real time in a complex environment, and provides key position and visual motion information for recovery operation.
By combining the KNN algorithm and the SORT algorithm, the method can play an advantage in the aspects of target identification and real-time tracking. The KNN algorithm provides accurate target recognition capability, and the SORT algorithm ensures real-time tracking of targets, so that accurate recognition and stable tracking of rocket sub-stages are realized. The combined algorithm can improve the performance and efficiency of the system and provide reliable support for rocket sublevel recovery.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the present application. As used herein, the singular is also intended to include the plural unless the context clearly indicates otherwise, and furthermore, it is to be understood that the terms "comprises" and/or "comprising" when used in this specification are taken to specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof.
It should be noted that the terms "first," "second," and the like in the description and the claims of the present application and the above figures are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that embodiments of the application described herein may be implemented in sequences other than those illustrated or otherwise described herein.
The above description is only of the preferred embodiments of the present application and is not intended to limit the present application, but various modifications and variations can be made to the present application by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the protection scope of the present application.