Small Target Tracking Research Articles

Infrared moving small target detection and tracking algorithm based on feature point matching

Infrared moving small target detection and tracking algorithm based on feature point matching

Small Target Recognition and Tracking Based on UAV Platform.

Target recognition and tracking based on multi-rotor UAVs have the advantages of low cost and high flexibility. It can monitor low-altitude targets with high intensity. It has great application prospects in national defense, military, and civil fields. The existing algorithms for aerial small target recognition and tracking have the disadvantages of slow speed, low accuracy, poor robustness, and insufficient intelligence. Aiming at the problems of existing algorithms, this paper first makes a lightweight improvement for the YOLOv4 network recognition algorithm suitable for small target recognition and tests it on the VisDrone dataset. The accuracy of the improved algorithm is increased by 1.5% and the speed is increased by 3.3 times. Then, by analyzing the response value, the KCF tracking situation is judged, and the template update of the adaptive learning rate is realized. When the tracking fails, the target is re-searched and tracked based on the recognition results and the similarity judgment. Finally, experiments are carried out on the multi-rotor UAV, and the adaptive zoom tracking strategy is designed to track pedestrians, cars, and UAVs. The results show that the proposed algorithm can achieve stable tracking of long-distance small targets.

Dim and Small Target Tracking Using an Improved Particle Filter Based on Adaptive Feature Fusion

Particle filters have been widely used in dim and small target tracking, which plays a significant role in navigation applications. However, their characteristics, such as difficulty of expressing features for dim and small targets and lack of particle diversity caused by resampling, lead to a considerable negative impact on tracking performance. In the present paper, we propose an improved resampling particle filter algorithm based on adaptive multi-feature fusion to address the drawbacks of particle filters for dim and small target tracking and improve the tracking performance. We first establish an observation model based on the adaptive fusion of the features of the weighted grayscale intensity, edge information, and wavelet transform. We then generate new particles based on residual resampling by combining the target position in the previous frame and the particles in the current frame with higher weights, with the tracking accuracy and particle diversity improving simultaneously. The experimental results demonstrate that our proposed method achieves a high tracking performance with a distance accuracy of 77.2% and a running speed of 106 fps, respectively, meaning that it will have a promising prospect in dim and small target tracking applications.

Real-time detection and tracking of infrared small targets based on grid fast density peaks searching and improved KCF

Improving the accuracy and meeting the real-time requirement of detection and tracking for infrared small target is always a research hot-spot. In this paper, we propose a target detection approach based on grid-based density peaks searching. First, in view of the small size of infrared target, Top-hat filter is introduced for preprocessing of raw infrared images, to enhance the target region. Then, the image is divided into image blocks by grid division. Compared to the original method of density peaks searching on each pixel, grid-based density peaks searching algorithm improve the detection speed on the basis of ensuring the detection accuracy. Finally, the small candidate target area is detected by the gray-scale area growth method, and the final small targets are segmented through a threshold. For the infrared small targets tracking, the improved correlation filtering tracking algorithm KCF is used. To address the issue of target loss in the KCF tracking process, we incorporate Kalman filtering into KCF. In the case of the target loss, Kalman filtering is adopted to predict the target’s trajectory. The combination of detection and tracking algorithm could significantly improve the accuracy of infrared small targets tracking, while the fusion of detection algorithm and the tracking algorithm of correlation filtering makes the real-time performance of the algorithm better and more efficient. Experimental results show that compared with several currently typical infrared small targets detection and tracking algorithms, the method proposed in this paper has better detection accuracy and speed.

Bernoulli track‐before‐detect smoothing for maritime radar

Detection and tracking of small targets in sea clutter using a high-resolution radar is a challenging problem as the sea surface constantly moves in a complex manner, with rough seas creating strong target-like returns. Recently, a Bernoulli track-before-detect (TBD) filter has been developed for an airborne scanning radar in the maritime domain. Its purpose is to detect and track short-exposure targets by using the radar in a fast scanning mode. The current paper investigates the potential improvement in detection and tracking performance of the Bernoulli TBD filter by incorporating a backward smoothing step. The developed algorithm, referred to as the Bernoulli TBD forward-backward smoother (FBS) for maritime radar, is making a delayed decision by taking into account multiple future scans of radar data. A numerical analysis of a joint detection and tracking error with a fixed-lag smoothing is presented, as a function of the number of delay lags, signal-to-interference ratio and the sea clutter spikiness.

Target tracking method based on interference detection

Considering the problems of similarity interference, partial occlusions, and changes in scale during target tracking, a target tracking method based on interference detection is proposed, which is an improvement over the Siamese fully convolutional classification and regression neural network (SiamCAR) approach. Under the proposed framework, the marginal distribution of the feature maps is used to determine the presence or absence of interferents. When interference is present in a scene, a motion vector composed of the predicted value obtained through a Kalman filter is used as the basis for target prediction. Experiments on the benchmark LaSOT dataset show that the proposed algorithm based on SiamCAR, which introduces motion features, achieves the best performance in videos with similar object interference, partial occlusions, fast motion, and small target tracking, as compared with the classical SiamCAR and other excellent target tracking algorithms.

Exploiting Doppler in Bernoulli Track-Before-Detect for a Scanning Maritime Radar

Detection and tracking of small targets in sea clutter using high-resolution radar is a challenging problem. Recently, a Bernoulli track-before-detect (TBD) filter has been developed for an airborne scanning radar in the maritime domain, with the purpose of detection and tracking of short-exposure targets by using a fast scanning mode (i.e., short dwell-time and noncoherent integration). This article investigates the potential benefits of coherent radar processing and exploitation of Doppler information in TBD, when the radar operates in a slower scanning mode (i.e., using longer dwell times). For this purpose, a new Bernoulli TBD filter is developed, which is capable of processing full three-dimensional (range-azimuth-Doppler) radar data. The amplitude of sea clutter is modeled using the K distribution with Doppler-dependent shape and scale parameters. Numerical results indicate that the new Bernoulli TBD outperforms the fast scanning TBD filter, at signal-to-interference ratios below 6 dB.

Infrared Small Object Tracking Based on Att-Siam Network

Aiming at the practical problems such as background clutter interference and occlusion during the tracking of small air targets by infrared imaging guided missiles, we propose an improved SiamFC infrared small target tracking algorithm. By mapping the original spatial data to a high-dimensional Hilbert space, the nonlinear mapping is implicit in the linear learner. This method has the advantages of learnability, efficient calculation, linearizability, and good generalization performance. The target tracking problem provides a new and effective approach with the actual needs of the application. We propose a Siamese web tracker (Att-Siam). Att-Siam fuses convolutional channel attention mechanism, stacked channel attention mechanism and spatial attention mechanism to improve tracking performance. Feature extraction of target objects is enhanced by fusing the channel attention mechanism and two convolution blocks at low convolutional layers. Compared with the benchmark algorithm SiamFC, the improved algorithm is tested on the infrared air weak and small target data set, and the tracking success rate and accuracy are increased by 32.3% and 20.9% respectively. The experimental results show that the proposed algorithm can adapt to complex and diverse infrared air scenes, and achieve effective and stable real-time tracking of small infrared air targets.

Siamese network with bidirectional feature pyramid for small target tracking

To address the tracking challenges such as weak feature expression ability of small targets and susceptibility to interference by similar objects in complex backgrounds, we use the principle of feature enhancement in the field of small target detection to redesign the backbone network of the siamese network tracker and propose a small target tracking algorithm based on a bidirectional feature pyramid fusion framework. The algorithm first constructs a deep feature pyramid with semantic and contextual information in the second half of the backbone network using successive deconvolution; then, considering that tracking also has a strong dependence on shallow information, we construct a shallow feature pyramid with location and spatial structure information in the first half of the backbone network as a complement to the deep feature pyramid, which is used to repair small target internal structure details and enhance its localization ability. Finally, the features processed from the deep and shallow pyramids are merged to construct a bidirectional pyramid fusion framework, while a self-attention mechanism is introduced to treat each type of information after fusion in a targeted manner, and the asymmetric convolution is used to lighten the fusion framework. The algorithm in this paper achieves a more advanced performance compared with existing algorithms for experiments on four publicly available datasets, GOT-10k, LaSOT, UAV123, and DTB70.

Anti-interference small target tracking from infrared dual waveband imagery

For infrared small target detection and tracking, the challenge mainly contains dim targets, cluttered background, and infrared decoys. In the face of the above difficulties, traditional tracking algorithms focus on utilizing information of a single waveband is difficult to achieve good performance. In this paper, a tracking algorithm, which combines the Kernelized Correlation Filter (KCF) with a detection model, is presented to cope with these challenges on dual waveband imagery. Firstly, the Side Window Box Filter (SWBF) is utilized to suppress the background edge in the wide-band image, which contains much more detail than that in narrow-band image. Therefore, several filters are obtained in the first wide-band image processed by the SWBF. Subsequently, fused features for the KCF algorithm can be computed via convolution operation. In order to solve the migration problem referring to decoys, a detection model based on multi-frame association and image correlation is introduced to refind the target. Finally, experimental results demonstrate that the presented algorithm outperforms several relative algorithms in the accuracy and robustness. Besides, the multi-source information leads to improve the performance of tracking methods.

A GM-CPHD Filtering Algorithm Assisted by Luminance Information

Optical tracking of small and weak targets in space, has problems such as a large and unknown number of targets, dense clutter, close measurement of neighboring targets, and false detection and missed detection. In this paper, a Gaussian mixture cardinalized probability hypothesis density (GM-CPHD) filtering algorithm assisted by luminance information is proposed. By establishing the luminance likelihood function, the luminance information is introduced into the CPHD filtering process, and the Gaussian mixture implementation method is given. The simulation results indicate that the proposed algorithm can effectively adapt to the dense clutter environment, and compared with the traditional GM-CPHD filtering algorithm, its tracking accuracy for multiple targets is significantly improved.

Infrared Small Target Tracking via Gaussian Curvature-Based Compressive Convolution Feature Extraction

The precision of infrared (IR) small target tracking is seriously limited due to lack of texture information and interference of background clutter. The key issue of robust tracking is to exploit generic feature representations of IR small targets under different types of background. In this letter, we present a new IR small target tracking method via compressive convolution feature (CCF) extraction. First, a Gaussian curvature-based feature map is calculated to suppress clutters so that the contrast between target and background can be obviously improved. Then, a three-layer compressive convolutional network, which consists of a simple layer, a compressive layer, and a complex layer, is designed to represent each candidate target by a CCF vector. Based on the proposed mechanism of feature extraction, a support vector machine (SVM) classifier with continuous probabilistic output is trained to compute the likelihood probability of each candidate. Finally, the long-term tracking for IR small target is implemented under the framework of the inverse sparse representation-based particle filter. Both qualitative and quantitative experiments based on real IR sequences verify that our method can achieve more satisfactory performances in terms of precision and robustness compared with other typical visual trackers.

Detection and tracking of infrared small target by jointly using SSD and pipeline filter

Infrared imaging has been an efficient anti-drone approach due to its low-cost, anti-interference and all-weather working characteristics. However, the detection of Unmanned Aerial Vehicle (UAV) through infrared camera is still a challenging issue because infrared targets in the field-of-view are usually small and lack of shape and texture features. In this paper, we propose an infrared small target detection and tracking method based on deep learning. We improve the network architecture of Single Shot MultiBox Detector (SSD) for infrared small target detection, called Single Shot MultiBox Detector for Small Target (SSD-ST), by dropping low-resolution layers and enhance high-resolution layer. In addition, in order to further reduce the false alarm rate and improve the precision, we also design an Adaptive Pipeline Filter (APF) based on the temporal correlation and motion information to correct the detection results. We have evaluated our method over a dataset with 16177 infrared images and 30 trajectories. The results show our method is more robust than traditional methods in complex scenes, and achieve a recall rate higher than 90% and a precision higher than 95%, which prove that our method can well complete the detection and tracking task of infrared small targets.

Low-Slow-Small Target Tracking Using Relocalization Module

With the gradual opening of airspace, tracking of noncooperative low-altitude slow-speed small size (LSS) targets is important for the maintenance of security. It is still a challenging problem, especially for complex scenarios and real-time constraints. In this letter, an efficient tracking by relocalization (TRL) framework is proposed for small flying object tracking, aiming to alleviate the issue of losing moving targets in a complex background. Our designed relocalization module consists of a feature-aggregated module and a global search module. On the one hand, a feature-aggregated module is integrated into the designed framework to increase the ability to locate small targets. On the other hand, a global search module is developed to update the tracking performance, which attempts to address missed targets in long-term small object tracking tasks. What needs to be declared is that the basic tracking module cooperates with the relocalization module we designed to achieve the tracking of small targets. Performance evaluation of two small-flying target data sets and comparison with several state-of-the-art approaches demonstrate the effectiveness of the proposed framework.

Markerless Tumor Tracking using Short Arc Digital Tomosynthesis With Fast-kV Switching Dual Energy Imaging

Markerless tumor tracking (MTT) using template matching is being considered for real-time lung tumor tracking using single energy (SE) and dual energy (DE) fluoroscopy. While MTT results using DE are generally superior vs. SE imaging, tracking small targets (< 10 mm) is challenging for both techniques. Short arc digital tomosynthesis (DTS), in which projections over a short (6 deg) sliding arc are combined to form a tomosynthesis image, may serve as a preprocessing technique to improve tracking. The goal of this study is to evaluate MTT with DTS using SE imaging and fast-kV switching DE imaging with the on-board imager (OBI) of a commercial linear accelerator. To evaluate these techniques, a motion phantom, consisting of a torso with embedded ribs and spine along with a cavity having lung-equivalent density, was used. A 5 mm simulated tumor was placed inside the lung equivalent compartment of the phantom. The simulated tumor was placed at three different static positions representing the two extremes and the mid-point of the respiratory cycle. Full rotation cone beam computed tomography (CBCT) acquisitions were obtained using SE and DE imaging (15 frames/sec) with a gantry speed of 6 deg/sec. DE images were processed using weighted logarithmic subtraction on consecutive high-low projections to produce soft tissue images. DTS image sequences were generated from both SE and DE image sequences. A template-based matching algorithm was used to track target motion on SE and DE fluoroscopy and SE- and DE-DTS images. Tumor tracking coordinates were evaluated against ground truth motion using the receiver operating characteristics (ROC) and root-mean-squared error (RMSE) on the combined results for all three tumor positions. The ROC area-under-the-curve (AUC) was the lowest for SE (0.82) and DE fluoroscopy (0.87). The use of SE- and DE-DTS increased the AUC values to 0.93 and 0.98, respectively. For a 95% specificity, the sensitivities for SE and DE fluoroscopy were 37% and 50%, respectively. The inclusion of SE-DTS increased this sensitivity to 81%. However, the highest sensitivity was observed for DE-DTS at 96%. With respect to local accuracy, the RMSE was evaluated based on a 95% specificity. For SE and DE fluoroscopy, the RMSE were 1.57 mm and 1.13 mm, respectively, while SE- and DE-DTS had RMSE of 0.58 mm and 0.54 mm, respectively. While DE fluoroscopy improves MTT vs. SE, both techniques have low sensitivity for small targets. The use of DTS significantly improves sensitivity and tracking accuracy vs. these fluoroscopic techniques. However, the most significant improvements were noted for DE-DTS, suggesting that this technique can be a useful preprocessing technique to make markerless tracking of small targets more robust and accurate.

UAV object tracking by background cues and aberrances response suppression mechanism

Real-time object tracking for unmanned aerial vehicles (UAVs) is an essential and challenging research topic for computer vision. However, the scenarios that UAVs deal with are complicated, and the UAV tracking targets are small. Therefore, general trackers often fail to take full advantage of their performance in UAV scenarios. In this paper, we propose a tracking method for UAV scenes, which utilizes background cues and aberrances response suppression mechanism to track in 4 degrees-of-freedom. Firstly, we consider the tracking task as a similarity measurement problem. In this study, we decompose this problem into two subproblems for optimization. Secondly, to alleviate the problem of small targets in UAV scenes, we utilize background cues fully. Also, to reduce interference by background information, we employ an aberrance response suppression mechanism. Then, to obtain accurate target state information, we introduce a logarithmic polar coordinate system. We perform phase correlation calculations in logarithmic polar coordinates to obtain the rotation and scale changes of the target. Finally, target states are obtained through response fusion, which includes displacement, scale, and rotation angle. Our approach is carried out in a large number of experiments on various UAV datasets, such as UAV123, DBT70, and UAVDT2019. Compared with the current advanced trackers, our method is superior on UAV small target tracking.

Improved GM-PHD filter based on threshold separation clusterer for space-based starry-sky background weak point target tracking

The movement of the background caused by the movement of a sensor platform has always been a major challenge in space-based infrared weak and small target tracking. In addition to clutter, false alarms, noise, and other disturbances, the most serious source of interference in space-based infrared weak target tracking comes from the stars. This kind of interference always exists and is large enough to be often mistakenly recognized as a new target by traditional algorithms. To reduce the influence of sensor platform movement and stellar interference on the target tracking of weak points in space, this paper respectively proposes a fast iterative closest point (ICP) registration algorithm for sparse target points and a threshold separation clustering algorithm. After pruning and merging using a Gaussian mixture probability hypothesis density (GM-PHD) filter-based algorithm, the threshold separation cluster separates the weak point targets by thresholds, and then clusters according to the Euclidean distance between the targets. Moreover, to prevent stars with special positions from being clustered into a single group, a dynamic weight extraction scheme is adopted to better distinguish stars in the group. We compared the proposed algorithm with the original GM-PHD, forward-backward smoothing (FBS), and N-scan approaches in starry-sky tracking scenarios simulated using the Tycho-2 catalog. Experimental results show that the proposed algorithm has better tracking accuracy and a lower rate of false detections.

Tracking of small infrared targets based on convolutional neural network

As an important application of computer vision, visual tracking has been studied more and more. However, there are still many problems to be solved for infrared target detection, especially for weak infrared target detection and tracking. In addition, convolutional neural networks are widely used in visual applications such as target detection, recognition, and tracking. Therefore, in this paper, an infrared weak target detection and tracking method is designed. Based on the criteria of visual saliency, the real target can estimate by calculating the Euclidean distance between the target and its neighbours. The detected position of the target is sent to a fully convolutional siamese network DeepCF for tracking. As we know, the factors such as lighting changes, scale changes and occlusion cause the tracking failure. Therefore, a tracking confidence T is proposed to judge whether the target is lost. The motion trajectory of the target is continuously estimated if it is lost. The experimental results show that the above method can effectively detect and track infrared weak targets.

Online Prostate-Specific Membrane Antigen and Positron Emission Tomography–Guided Radiation Therapy for Oligometastatic Prostate Cancer

PurposeStereotactic ablative radiation therapy (SABR) for oligometastatic prostate cancer (OMPC) may improve clinical outcomes, but current challenges in intrafraction tracking of multiple small targets limits treatment accuracy. A biology-guided radiation therapy (BgRT) delivery system incorporating positron emission tomography (PET) detectors is being developed to use radiotracer uptake as a biologic fiducial for intrafraction tumor tracking to improve geometric accuracy. This study simulates prostate-specific membrane antigen (PSMA)-directed BgRT using a cohort from our phase II randomized trial of SABR in men with recurrent hormone sensitive OMPC and compares dose distributions to clinical SABR (CSABR). Methods and MaterialsA research treatment planning system (RTPS) was used to replan 15 patients imaged with PSMA-targeted 18F-DCFPyL PET/computed tomography and previously treated with CSABR using conventional linear accelerators (linacs). The RTPS models a prototype ring-mounted linac incorporating PET and kilo-voltage computed tomography imaging subsystems and can be used to optimize BgRT plans, as well as research SABR (RSABR) plans, which use the prototype linac without radiotracer guidance. CSABR, RSABR, and BgRT plans were compared in terms of maximum planning target volume (PTV) dose (Dmax), mean dose to proximal organs at risk (DOAR), conformity index, as well as voxel-wise correlation of dose with PET specific uptake values to investigate possible dose-painting effects. ResultsRSABR and BgRT plans resulted in mean ± standard deviation increases in Dmax of 4 ± 11% (P = .21) and 18 ± 15% (P < .001) and reductions in DOAR of –20 ± 19% (P <.001) and –10 ± 19% (P = .02) compared with CSABR. Similar target coverage was maintained with conformity indices of 0.81 ± 0.04 (P < .001) and 0.72 ± 0.08 (P = .44) for RSABR and BgRT compared with 0.74 ± 0.08 for CSABR. Dose and log (specific uptake values) had Pearson correlation coefficients of 0.10 (CSABR), 0.16 (RSABR), and 0.31 (BgRT). ConclusionsBgRT plans provided similar PTV coverage and conformity compared with CSABR while incorporating underlying PET activity. These results demonstrate feasibility of BgRT optimization enabling online PSMA-targeted, PET-based tracked dose delivery for OMPC.

基于嵌入式平台的低时间复杂度目标跟踪算法

Aimed at the application background of embedded platform which is often limited in computing power, this paper proposed a low time complexity target tracking algorithm, CTSTC algorithm, which was suitable for complex scenes. The algorithm consisted of two parts:the main part constituted by the spatio-temporal context target tracking based on adaptive update of model and the aided target location part constituted by compressive tracking based on adaptive update of model. When the results of spatio-temporal context tracking were unreliable, the aided location part was activated. If the results of aided location were reliable, the aided location result was used to correct the spatio-temporal context tracking part. The running speed of the algorithm was close to that of the spatio-temporal context learning algorithm (STC). The test on I5CPU can reach 1 577 frames per second, which was much faster than other commonly used algorithms. It was a very fast target tracking algorithm, but the robustness of the algorithm in complex environments was improved. Using OTB2013 data set to test, compared with STC algorithm, CTSTC accuracy increased by 12.8%, success rate increased by 27.5%. The algorithm is tested on a small target tracking system with DM6437 as the core, which can achieve real-time stable tracking.