The invention discloses a tracking-learning-detection (TLD)-based video object tracking method, which comprises the following steps of: (1) when a start frame of an object is tracked, generating an image sub-window according to given position and size information, and training a detector; (2) estimating the area of the object in a current frame according to the position and size information of the object in a previous frame by using a tracker; (3) detecting the current frame by using the detector, and finding all possible objects in the current frame; and (4) performing fusion processing on results of the tracker and the detector, judging whether the object exists in the current frame or not, returning to the step (3) to process the next frame if the object does not exist in the current frame, judging whether a tracking trajectory of the current frame is effective or not if the object exists in the current frame, giving the position of the object, returning to the step (2) to process the next frame if the tracking trajectory of the current frame is ineffective, finishing the online updating of the detector if the tracking trajectory of the current frame is effective, and returning to the step (2) to process the next frame. According to the method, the conventional TLD algorithm is improved to obtain a video object tracking algorithm which is more ideal than the TLD algorithm.

Tracking-learning-detection (TLD)-based video object tracking method

The invention discloses a face age estimation method performing measurement learning based on a convolutional neural network. The method comprises the overall steps that a dataset is constructed; thedataset is divided into a training set and a verification set; paired construction is performed on mini-batches on a network input layer, and then the mini-batches are sent into two twin networks fortraining; a VGG-16 network is constructed; network training is performed; softmax loss and revised contrastive loss are jointly used as supervisory signals to perform network adjustment; network evaluation is performed; and the finally estimated age is a maximum probability corresponding category obtained on a softmax layer. According to the method, deep learning and measurement learning are combined; by introducing measurement learning, the distinction degree of a feature space is higher, and therefore the robustness of an age estimation algorithm is higher; and deep learning is utilized to combine a feature extraction task and an objective function optimization task, end-to-end training is realized for the whole task, and good performance can be obtained when the method is applied to a public dataset.

Face age estimation method performing measurement learning based on convolutional neural network

The invention discloses a method for training a speaker recognition network model. The method comprises the steps of acquiring the ternary group of i utterance from the training data set, and inputting the ternary group of the i utterance into a convolutional neural network, the convolutional neural network carrying out characteristic extraction on the ternary group of the i utteranceto obtain a ternary group of the i utterance characteristic and inputting into the linear neural network; inputting the first identity vector information of the first speaker and the second identity vector information of the second speaker into a linear neural network for fusion processing, so as to obtain the fused ternary group of the i utterance; and calculating a ternary group loss according to the ternary group of the i utterance so as to adjust the network model. The speaker recognition network model obtained by the embodiment of the invention can realize the recognition function of speakers more accurately, and has the lowest error rate.

Method for training speaker recognition network model, speaker recognition method and system

The invention relates to a heterogeneous image matching method based on deep learning A heterogeneous image block data set is made; image preprocessing is carried out; an image block feature map is obtained; a feature vector is obtained based on the feature map; feature map fusion and normalization are carried out; an image matching network is trained; and a matching probability is predicted Therefore, a problem of over-fitting of heterogeneous image block matching in the prior art is solved; the performance of the network is improved substantially; the training efficiency of the network isenhanced; and the robustness of the network is enhanced The heterogeneous image matching method can be applied to fields of heterogeneous image registration, image tracking, multi-view reconstructionand the like

Heterogeneous image matching method based on deep learning

The invention discloses a CNN and Kalman filtering based moving target tracking method and system based on region selection. According to a convolutional neural network based target tracking algorithmprovided by the invention, macroscopic motion information of a target is added as a reference in a tracking process; a Kalman filter is used to perform feasibility pre-determination on a tracking target motion path; the optimal estimation of the target point at an image plane position obtained by the filter is used as a starting point of convolutional neural network target recognition in each frame, and the target position is globally pre-estimated; the identification detection problem under different background interference conditions is analyzed; the problems of long training period, largecalculation amount and high false positive rate of the training model are solved; and the tracking performance is improved.

CNN and Kalman filtering based moving target tracking method and system based on region selection

The invention relates to a target tracking method based on a convolution neural network. The main technical characteristics comprise: building a convolution neural network framework suitable to track tasks; under constraint of a center contrast loss function, performing parameter training on the built convolution neural network; tuning a trained model structure, and performing on-line tracking. The method is rational in design, uses the convolution neural network as a structural basis, and focuses on within cluster variation among different objects, and can preferably cope with conditions of mixed background and interference of similar targets, and has good robustness and high accuracy.

Target tracking method based on convolution neural network

Visual tracking is a significant but challenging field in computer vision. Although considerable progress has been made in recent years, robust tracking in complicated scenes remains an open problem. Trackers get confused easily when similar objects appear or heavy clutter occurs due to indistinguishable features. In this work, a more effective feature extraction method based on convolutional neural network (CNN) is proposed. Different from conventional CNN models, this method applies a contrastive loss function to a single branch network, and centralization is employed to obtain more discriminative information. In addition, appropriate model update is used to capture transformation in object appearance. Quantitative experimental results on various video sequences demonstrate the superior performance of the proposed method in comparison with other state-of-the-art trackers. Complementary experiments are also conducted to validate some arguments.

Center contrastive loss regularized CNN for tracking

Numerous trackers have been proposed in recent years with considerable success. But few trackers can cope with all scenarios without failures. It is very difficult to design a tracker robust enough to keep off tracking failure. As failure is inevitable, we propose a framework to correct tracker, verify failure, predict object position and re-detect object. The original model of the first frame is used to correct the tracker. Then, the confidence of tracking is used to verify tracking failure. Finally, the spatial context is used to predict the object location when the object is invisible and re-detect object after tracking failure. A variety of comparative experiments validate the superiority of our tracker in comparison against other trackers.

Recovery from tracking failure

Visual tracking has attracted much research due to its importance in many practical applications such as video surveillance, vehicle navigation and augmented reality. Although existing trackers have made some success under various scenarios, long-term robust tracking in real-world scenes remains an open problem. Trackers are easily drifting away without external re-initialization due to drastic deformation, heavy occlusion and complex environments. We propose a framework to correct tracker by memorizing diverse states of the object. A memory pool of object state is constructed first, then maximum posterior probability is employed to estimate whether the object is present in the view and a best fit model is picked out from the pool for current frame. The qualitative and quantitative experimental results on various long video sequences demonstrate the superior performance of our method in comparison with other long-term trackers.

Correcting the tracker with memories

Visual tracking is a significant but challenging field in computer vision. Long-term robust tracking in unconstrained environments is insurmountable due to arbitrary deformation, occlusion and background clutters. In this paper, we propose a framework to integrate couplers and historical models into a discriminative classifier for robust long-term tracking. Model updating inevitably introduces errors into the tracker, and the appearance model cannot match the object well. A best fitted model is retrieved from historical models to correct the tracker. Couplers, which are key points around the object that tightly coupled with the target, are detected to predict the position of the target when the tracker is fully occluded or distracted by other similar objects. Both quantitative and qualitative experimental results validate the superiority of our tracker over other state-of-the-art methods.

Tracking with the support of couplers and historical models

Handle everyday research tasks with reliable, citation-backed results

Your personal Research Agent to handle research tasks with citation-backed results

Popular Tasks used by Researchers

How can I help with your research?

Meet SciSpace

Get more enhanced response by uploading the PDFs you want me to reference.

No relevant PDFs in your library

SciSpace is the AI research assistant for academics. Run systematic literature reviews on 280M+ papers, and write papers with cited sources. Trusted by 1M+ students, PhDs & researchers.

SciSpace | AI for Research

Analyze PDFs

Code & Manuscripts

Funding & Grants

Literature & Patents

Medical & Clinical Data

Systematic Review

Visualize & Present

Web & Data

Build a Google Scholar-like website for your research.

Build a website

Create charts and images for your research

Create a Chart

Write a paper for submission to a journal

Draft a manuscript

Patent Search

Design eye-catching scientific posters in minutes.

Scientific Poster Generation

Systematic Literature Review

One task is running at the moment. Your messages will be shown right after.

Drag and drop or click here to browse

Loved by <highlight>1 million+</highlight> researchers

Extract a list of specific topics and their sources from unstructured text

Topics

Compare and analyze relevant papers that matches with your search

Papers

Get insights from PDFs and bookmarked papers from your library

My library

Recent searches

Try searching for:

Catch AI-generated content in scholarly and non-scholarly content

{ai} Detector

Ai Writer

Get PDF Summaries, highlighted text explanations 

Chat with PDF

Effortlessly create in-text citations and bibliographies in APA and 2,500 other formats

Citation generator

Get explanations, summaries, and answers on academic papers

Ease up your research workflow with {scispace}'s cohort of exciting AI tools

Elevate your academic writing skills and convey your ideas the way you want

Paraphraser

Explore our range of reading and writing tools

Your file is being prepared and should be ready in a few minutes. If it's a large file, it might take a bit longer. You can close this window, and we'll email you the file when it's done.

You have reached a maximum limit of <strong>{limit}</strong> columns in the table. Remove at least <strong>1</strong> column to add or create another one.

Ningning Li

Author Tools

Chat about Author