Bishopdissing1933
Siamese networks are prevalent in visual tracking because of the efficient localization. The networks take both a search patch and a target template as inputs where the target template is usually from the initial frame. Meanwhile, Siamese trackers do not update network parameters online for real-time efficiency. The fixed target template and CNN parameters make Siamese trackers not effective to capture target appearance variations. In this paper, we propose a template updating method via reinforcement learning for Siamese regression trackers. We collect a series of templates and learn to maintain them based on an actor-critic framework. N6F11 order Among this framework, the actor network that is trained by deep reinforcement learning effectively updates the templates based on the tracking result on each frame. Besides the target template, we update the Siamese regression tracker online to adapt to target appearance variations. The experimental results on the standard benchmarks show the effectiveness of both template and network updating. The proposed tracker SiamRTU performs favorably against state-of-the-art approaches.In order to effectively and flexibly control acoustic pattern, an efficient optimization design method of acoustic liquid lens (ALL) is developed by the frame of particle swarm optimization (PSO) algorithm. The ALL is composed of ethanol and dimethicone, and its parameters include ethanol concentration (EC), volume fraction of dimethicone (VFD) and total volume (TV). Based on the established finite element model and orthogonal design method, the data of acoustic pattern and ALL can be obtained by using COMSOL Multiphysics. Based on the simulation data, the neural network models are constructed to characterize the relationship between the parameters of ALL and the performance of acoustic pattern. The optimization design criteria of ALL are constructed based on the performance parameters of acoustic pattern, including focal distance (FD), transverse resolution (TR) and longitudinal resolution (LR). Based on the optimization criteria, the modified PSO algorithm is utilized to optimize the design parameters of ALL in the developed method. According to desired FD, TR and LR of acoustic pattern (20, 1 and 17 mm), the optimized EC, VFD and TV of ALL are about 0.838, 0.165 and 164.4 μL. The performance parameters of acoustic pattern verified by simulation and experiments agree with the desired ones. In addition, using 6 MHz ultrasonic transducer with the optimized ALL, the ultrasonic imaging of tungsten wires and porcine eyeball further demonstrates the effectiveness and feasibility of the developed method.This paper proposes a mixed low-rank approximation and second-order tensor-based total variation (LRSOTTV) approach for the super-resolution and denoising of retinal optical coherence tomography (OCT) images through effective utilization of nonlocal spatial correlations and local smoothness properties. OCT imaging relies on interferometry, which explains why OCT images suffer from a high level of noise. In addition, data subsampling is conducted during OCT A-scan and B-scan acquisition. Therefore, using effective super-resolution algorithms is necessary for reconstructing high-resolution clean OCT images. In this paper, a low-rank regularization approach is proposed for exploiting nonlocal self-similarity prior to OCT image reconstruction. To benefit from the advantages of the redundancy of multi-slice OCT data, we construct a third-order tensor by extracting the nonlocal similar three-dimensional blocks and grouping them by applying the k-nearest-neighbor method. Next, the nuclear norm is used as a regularization term to shrink the singular values of the constructed tensor in the non-local correlation direction. Further, the regularization approaches of the first-order tensor-based total variation (FOTTV) and SOTTV are proposed for better preservation of retinal layers and suppression of artifacts in OCT images. The alternative direction method of multipliers (ADMM) technique is then used to solve the resulting optimization problem. Our experiments show that integrating SOTTV instead of FOTTV into a low-rank approximation model can achieve noticeably improved results. Our experimental results on the denoising and super-resolution of OCT images demonstrate that the proposed model can provide images whose numerical and visual qualities are higher than those obtained by using state-of-the-art methods.Dynamic optical imaging of retinal hemodynamics is a rapidly evolving technique in vision and eye-disease research. Video-recording, which may be readily accessible and affordable, captures several distinct functional phenomena such as the spontaneous venous pulsations (SVP) of central vein or local arterial blood supply etc. These phenomena display specific dynamic patterns that have been detected using manual or semi-automated methods. We propose a pioneering concept in retina video-imaging using blind source separation (BSS) serving as an automated localizer of distinct areas with temporally synchronized hemodynamics. The feasibility of BSS techniques (such as spatial principal component analysis and spatial independent component analysis) and K-means based post-processing method were successfully tested on the monocular and binocular video-ophthalmoscopic (VO) recordings of optic nerve head (ONH) in healthy subjects. BSSs automatically detected three spatially distinct reproducible areas, i.e. SVP, optic cup pulsations (OCP) that included areas of larger vessels in the nasal part of ONH, and "other" pulsations (OP). The K-means post-processing reduced a spike noise from the patterns' dynamics while high linear dependence between the non-filtered and post-processed signals was preserved. Although the dynamics of all patterns were heart rate related, the morphology analysis demonstrated significant phase shifts between SVP and OCP, and between SVP and OP. In addition, we detected low frequency oscillations that may represent respiratory-induced effects in time-courses of the VO recordings.The performance of most the clustering methods hinges on the used pairwise affinity, which is usually denoted by a similarity matrix. However, the pairwise similarity is notoriously known for its venerability of noise contamination or the imbalance in samples or features, and thus hinders accurate clustering. To tackle this issue, we propose to use information among samples to boost the clustering performance. We proved that a simplified similarity for pairs, denoted by a fourth order tensor, equals to the Kronecker product of pairwise similarity matrices under decomposable assumption, or provide complementary information for which the pairwise similarity missed under indecomposable assumption. Then a high order similarity matrix is obtained from the tensor similarity via eigenvalue decomposition. The high order similarity capturing spatial information serves as a robust complement for the pairwise similarity. It is further integrated with the popular pairwise similarity, named by IPS2, to boost the clustering performance.