Search

In this paper, we target the “precision visual tracking” problem, which is the precise tracking of a target point on a moving object. Compared with the abundant efforts in object-level tracking, which aims at accurately predicting the bounding box (or the mask) of the object, much less attention has been drawn to precision visual tracking. To this end, we present a template tracking framework that consists of two parts, template matching and template updating. For template matching, we trained a deep architecture to directly estimate the projective transformation that deforms the template to the search image. For template updating, we came up with a systematic strategy to update the initial and new templates. To avoid drift build-ups, we incorporate a fast-running dense correspondence matching module into the template update step. The proposed method was extensively tested on both synthetic and real data. To generate the synthetic data, we created a dataset of 480 image sequences. Each image sequence comes with the trajectory ground truth of a target point moving across all the frames. We compared the proposed method with six comparative approaches on this dataset. The tracking accuracy achieved by the proposed method was as follows: around 44% of the tracking errors were less than one pixel, about 78% of them were less than three pixels, and only about 14% of them exceeded five pixels. The proposed method is fast, running at 14 frames-per-second (fps) on [Formula: see text] image sequences on a workstation equipped with a Nvidia RTX 2080Ti graphic card. Qualitative results also show that the proposed method is applicable to real-world image sequences. The related code, pre-trained models, and the test data will be made publicly available at https://github.com/XM-Peng/Precision-Visual-Tracking/.

In this paper, we propose a new visual tracking method in light of salience information and deep learning. Salience detection is used to exploit features with salient information of the image. Complicated representations of image features can be gained by the function of every layer in convolution neural network (CNN). The characteristic of biology vision in attention-based salience is similar to the neuroscience features of convolution neural network. This motivates us to improve the representation ability of CNN with functions of salience detection. We adopt the fully-convolution networks (FCNs) to perform salience detection. We take parts of the network structure to perform salience extraction, which promotes the classification ability of the model. The network we propose shows great performance in tracking with the salient information. Compared with other excellent algorithms, our algorithm can track the target better in the open tracking datasets. We realize the 0.5592 accuracy on visual object tracking 2015 (VOT15) dataset. For unmanned aerial vehicle 123 (UAV123) dataset, the precision and success rate of our tracker is 0.710 and 0.429.

Neural network pruning, an important method to reduce the computational complexity of deep models, can be well applied to devices with limited resources. However, most current methods focus on some kind of information about the filter itself to prune the network, rarely exploring the relationship between the feature maps and the filters. In this paper, two novel pruning methods are proposed. First, a new pruning method is proposed, which reflects the importance of filters by exploring the information in the feature maps. Based on the premise that the more information there is, more important the feature map is, the information entropy of feature maps is used to measure information, which is used to evaluate the importance of each filter in the current layer. Further, normalization is used to realize cross layer comparison. As a result, based on the method mentioned above, the network structure is efficiently pruned while its performance is well reserved. Second, we proposed a parallel pruning method using the combination of our pruning method above and slimming pruning method which has better results in terms of computational cost. Our methods perform better in terms of accuracy, parameters, and FLOPs compared to most advanced methods. On ImageNet, it is achieved 72.02% top1 accuracy for ResNet50 with merely 11.41M parameters and 1.12B FLOPs.For DenseNet40, it is obtained 94.04% accuracy with only 0.38M parameters and 110.72M FLOPs on CIFAR10, and our parallel pruning method makes the parameters and FLOPs are just 0.37M and 100.12M, respectively, with little loss of accuracy.

Turbulence-degraded image restoration is an important part in detection system which based on image. Most of current researches on turbulence-degraded image were focus on getting perfect image and not very care about processing speed. They are not acceptable when they apply on a real-time detection system. In order to restore degraded image clearly and rapidly, in this paper we introduce an efficiency restoration method for turbulence-degraded image base on improved SeDDaRA (self-deconvolving data reconstruction algorithm) method. SeDDaRA transform the image data form space field to spectrum field and smooth image's spectrum data and use a power law relation applied to the smoothed spectrum data to extract a filter function. This filter function can be used to restore and enhance higher-frequency content and get the system's Point Spread Function (PSF). The PSF can be used for deconvolution filter such as Winner and nonnegative least squares to restore the image. There are three major contributions in this paper. Firstly, we add a pre-denoise process to remove the noise which introduced by system such as period noise and Gauss noise. This step can significant improve the restore image's quality. Secondly we use an optimum method to extract the filter function which responded to PSF. The method, based on spectrum's power law characteristic, only need compute 8-direction date of the whole data to get the parameter. Compared with normal SeDDaRA method which need compute all data in spectrum field the new method can significant reduce the compute complication. Thirdly we utilize image's inherent characteristic and introduce a novel method to estimation deconvolution filter's SNR. The accurate SNR can efficiently improve the restoration quality. Compared with other restoration method, our method is noniterative and requires only that the point-spread function be space invariant and the transfer function be real. These mean that our method can work efficiently and requires little knowledge of the original data or the degradation. Experiments on real turbulence-degraded image indicate that the proposed method is very fast and can get quality restore image, which demonstrates the feasibility and validity of the proposed method.

This paper presents an iterative pose estimation method on the basis of point correspondences, which are composed of 3D coordinates of feature points under object reference frame and their 2D projective coordinates under image reference frame. The proposed method decomposes the pose estimation into two steps. Firstly, the 3D coordinates of the feature points under camera reference frame are estimated iteratively by Gauss-Newton method. In this process, the variables are defined by the lengths of the vectors from the focus point of camera to the feature points; meanwhile, several novel constraints are constructed by a set of error functions built out of the inner angles and areas of the triangles formed by three arbitrary non-collinear feature points, because they can describe the shape of object uniquely and completely. Secondly, by using Gauss-Newton method again, the rotation angles (i.e., pitch, yaw, and roll) and 3D translation of the object are estimated from the 3D coordinates of the feature points under camera reference frame obtained in the first step. Experiments involving synthetic data as well as real data indicate that the proposed method is more accurate and no less fast than the previous method.