Search

A new video based multi behavior dataset for cows, CBVD-5, is introduced in this paper. The dataset includes five cow behaviors: standing, lying down, foraging,rumination and drinking. The dataset comprises 107 cows from the entire barn, maintaining an 80% stocking density. Monitoring occurred over 96 h for these 20-month-old cows, considering varying light conditions and nighttime data to ensure standardization and inclusivity.The dataset consists of ranch monitoring footage collected by seven cameras, including 687 video segment samples and 206,100 image samples, covering five daily behaviors of cows. The data collection process entailed the deployment of cameras, hard drives, software, and servers for storage. Data annotation was conducted using the VIA web tool, leveraging the video expertise of pertinent professionals. The annotation coordinates and category labels of each individual cow in the image, as well as the generated configuration file, are also saved in the dataset. With this dataset,we propose a slowfast cow multi behavior recognition model based on video sequences as the baseline evaluation model. The experimental results show that the model can effectively learn corresponding category labels from the behavior type data of the dataset, with an error rate of 21.28% on the test set. In addition to cow behavior recognition, the dataset can also be used for cow target detection, and so on.The CBVD-5 dataset significantly influences dairy cow behavior recognition, advancing research, enriching data resources, standardizing datasets, enhancing dairy cow health and welfare monitoring, and fostering agricultural intelligence development. Additionally, it serves educational and training needs, supporting research and practical applications in related fields. The dataset will be made freely available to researchers world-wide. [ABSTRACT FROM AUTHOR]

Abstract A new video based multi behavior dataset for cows, CBVD-5, is introduced in this paper. The dataset includes five cow behaviors: standing, lying down, foraging,rumination and drinking. The dataset comprises 107 cows from the entire barn, maintaining an 80% stocking density. Monitoring occurred over 96 h for these 20-month-old cows, considering varying light conditions and nighttime data to ensure standardization and inclusivity.The dataset consists of ranch monitoring footage collected by seven cameras, including 687 video segment samples and 206,100 image samples, covering five daily behaviors of cows. The data collection process entailed the deployment of cameras, hard drives, software, and servers for storage. Data annotation was conducted using the VIA web tool, leveraging the video expertise of pertinent professionals. The annotation coordinates and category labels of each individual cow in the image, as well as the generated configuration file, are also saved in the dataset. With this dataset,we propose a slowfast cow multi behavior recognition model based on video sequences as the baseline evaluation model. The experimental results show that the model can effectively learn corresponding category labels from the behavior type data of the dataset, with an error rate of 21.28% on the test set. In addition to cow behavior recognition, the dataset can also be used for cow target detection, and so on.The CBVD-5 dataset significantly influences dairy cow behavior recognition, advancing research, enriching data resources, standardizing datasets, enhancing dairy cow health and welfare monitoring, and fostering agricultural intelligence development. Additionally, it serves educational and training needs, supporting research and practical applications in related fields. The dataset will be made freely available to researchers world-wide.

Background. Visual non-destructive testing of the inner surface of pipes is an important aspect in their production and operation. A defect detected and corrected in a timely manner can significantly reduce the number of defects in production and prevent various emergency incidents during operation. Formation of a complete panoramic image of the inner surface of pipes suitable for quality analysis is an urgent and sought-after task that can be solved using computer vision systems. Aim. This work is the research and development of television methods for forming a complete panoramic image of the inner surface of a pipe, which can be analyzed to search for defects. Methods. To form a cylindrical panoramic image, mathematical models for the formation of an equidistant projection of spherical images obtained using a fisheye lens were used. For high-quality stitching of the resulting frames, digital image processing methods were used, including brightness and contrast transformations, and searching for special points using the MSER algorithm. Theoretical results are verified by full-scale modeling. Results. The result of this work is an algorithm for stitching frames of a video sequence generated by a television camera with a fisheye optical system, uniformly moved along the longitudinal axis of the pipe, into a single panoramic image of the internal surface. Conclusion. The algorithm ensures the formation of a high-quality image of a full panorama of the inner surface of the pipes with the absence of brightness artifacts.

In recent years, there has been a lot of interest in the estimation of missing frames in video sequences because of its usage in several critical applications such as medical imaging, manufacturing, security systems, etc. The present study employs the Kalman filtering approach to suggest a technique for estimating missing frames within a video sequence and forecasting forthcoming frames. The Kalman filter model proposed in this study employs a pixel-by-pixel approach to leverage information from both preceding and subsequent video frames for the purpose of predicting the missing frame. The experimental findings demonstrate the effectiveness of the proposed technique. [ABSTRACT FROM AUTHOR]

Video imaging in open environments faces several problems such as the haze problem. Haze has a degradation effect on images. To enhance the video quality, we use a dual-transmission-map dehazing algorithm. We work on two video sequences: a Near Infrared (NIR) sequence and a visible sequence. We modify this dual-transmission-map dehazing algorithm by applying pre-processing on frames before the dehazing process to enhance the frame quality. All frames may have some noise or limited dynamic range due to sensor measurement errors that can be amplified in the haze removal process if ignored. We use different types of image enhancement techniques to either remove noise or control the dynamic range before the dehazing process. The dehazing algorithm can be implemented using a dual-transmission map depending on the Dark Channel Prior (DCP) method. Gradient-domain guided filtering is used to refine the two obtained transmission maps to obtain the best results for the dehazed frames. We investigate the effect of a regularization parameter ω on the visible and NIR dehazed frames without and with enhancement techniques. Results for visible video are more obvious than those for NIR video, because hazy visible frames have a large amount of haze compared to hazy NIR frames. We use the Peak Signal-to-Noise Ratio (PSNR) and correlation between hazy and dehazed frames after the dehazing process as evaluation metrics for the proposed algorithm. The proposed algorithm provides good results in the enhancement of two types of video as proved by the spectral entropy of dehazed frames. [ABSTRACT FROM AUTHOR]

Video in an outdoor environment suffers from the haze problem due to smoke, dust, and other particles in the atmosphere. The quality of video under these atmospheric conditions is clearly/dramatically degraded, leading to poor contrast and information leakage. The proposed algorithm in this paper depends on pre-processing of video frames prior to the dehazing process. Different enhancement techniques are used to remove noise and control the dynamic range, because each frame has some noise due to sensor measurement errors. This noise might be magnified during the haze removal process, if totally neglected. Previous dehazing algorithms enhance the video contrast and reduce frame degradation. We use optimized dehazing algorithms, namely a dual-transmission-map dehazing algorithm, and a recursive Deep Residual Learning (DRL) network, as dehazing tools. We modify the previous dehazing algorithms to generate new algorithms to be suitable not only for visible frames but also for Near Infrared (NIR) frames. In the optimized dehazing algorithms, the haze effect is reduced, according to the hazing model. In the dual-transmission-map dehazing algorithm, the effect of attenuation parameter R and attenuation weight U on the dehazed frames is investigated. A dual-transmission-map algorithm, depending on the Dark Channel Prior (DCP) technique, is studied. We study the effect of a regularization parameter Ω on the visible and NIR dehazed frames without and with enhancement techniques. For the recursive deep residual learning algorithm, we increase the number of iterations in the DRL network from three iterations in the traditional DRL algorithm without denoising techniques to nine iterations, in order to explore the impact of increasing the number of iterations on the output dehazed frames. Since elapsed time grows with the number of iterations, we terminate the operation after the ninth iteration. A comparison between traditional algorithms and the proposed dehazing algorithms is introduced for different types of video considering five frames. In addition, a comparison between previous algorithms and the proposed dehazing algorithms is introduced on real-world hazy images. We use correlation, Peak Signal-to-Noise Ratio (PSNR) between input hazy frame and dehazed one, histogram, and spectral entropy of hazy and dehazed frames as metrics for the proposed algorithms. We conclude that the best dehazing algorithm for the visible videos is the dual-transmission-map dehazing algorithm with enhancement. On the other hand, the best one for NIR videos is the optimized dehazing algorithm with enhancement. For the visible (riverside.avi) video, the enhancement score of the proposed dual-transmission-map dehazing algorithm is 14.74%. It is increased by 9.65% compared to the original dual-transmission-map dehazing algorithm. In contrary, the DRL network achieves a 9% enhancement, which is the lowest one due to the need for a huge training dataset. For NIR video, the enhancement scores of the three different proposed dehazing algorithms, the optimized, the dual-transmission-map, and the DRL network, are 22.28%, 21.57%, and 4.6%, respectively. [ABSTRACT FROM AUTHOR]

This article introduces the phenomenon of videopoetry as a hybrid product of mass media whose popularity is based on intermediality, i.e., the cumulative effect on different perception channels. Videopoetry is a productive form of verbal creativity in the contemporary media culture with its active reception of art. The research featured poems by W. B. Yeats, T. S. Eliot, and W. H. Auden presented as videos and the way they respond to someone else's poetic word. The authors analyzed 15 videos by comparing the original text and the video sequence in line with the method developed by N. V. Barkovskaya and A. A. Zhitenev. The analysis revealed several options for relaying a poetic work as a music video. Three videos provided a direct illustration of the source text, suggesting a complete or partial visual duplication of the original poetic imagery. Five videos offered an indirect illustration of the source text by using associative images in relation to the central images of the poem. Five videos gave a minimal illustration: the picture did not dominate the text of the poem, but its choice implied a certain interpretation. Two videos featured the video maker as a reciter. The video makers did not try to transform the poetic text but used the video sequence as a way to enter into a dialogue with the original poem or resorted to indirect illustration to generate occasional meanings. Thus, video makers keep the original text unchanged and see the video sequence and musical accompaniment as their responsibility but maintain a dialogue between the original text and its game reinterpretation.

Using the Kalman filtering technique, we propose a novel method for estimating the missing video frames to monitor the activities inside the control room of a nuclear power plant (NPP). The purpose of this study is to reinforce the existing security and safety procedures in the control room of an NPP. The NPP control room serves as the nervous system of the plant, with instrumentation and control systems used to monitor and control critical plant parameters. Because the safety and security of the NPP control room are critical, it must be monitored closely by security cameras in order to assess and reduce the onset of any incidents and accidents that could adversely impact the safety of the NPP. However, for a variety of technical and administrative reasons, continuous monitoring may be interrupted. Because of the interruption, one or more frames of the video may be distorted or missing, making it difficult to identify the activity during this time period. This could endanger overall safety. The demonstrated Kalman filter model estimates the value of the missing frame pixel-by-pixel using information from the frame that occurred in the video sequence before it and the frame that will occur in the video sequence after it. The results of the experiment provide evidence of the effectiveness of the algorithm.

To minimize any impediments in real-time Internet of Things (IoT)-enabled robotics applications, this study demonstrated how to build and deploy a revolutionary framework using computer vision and deep learning. In contrast to robotic path planning algorithms based on geolocation. We focus on sensor-captured streams/images and geographical information to enable the Internet of Robotic Things (IoRT) to evolve. The application will collect real-time data from moving robotics at various situations and intervals and use it for research projects. The data collected in videos/image forms are delivered in the robotics application using visual sensor nodes. In this study, anticipating moving robot moves automatically early on can aid in issuing commands to monitor and regulate robots' future activities before they occur. To do so, we propose the framework using efficient computer vision techniques and a deep learning classifier. The computer vision methods are designed for frame quality improvement, object segmentation, and feature estimation. The Long-Term Short Memory (LSTM) classifier detects robot motions automatically from initial sequential features. We mainly designed the proposed model using an LSTM classifier to perform the earlier prediction from the initial sequential features of partial video frames and to overcome the problems of exploding and vanishing gradients. LSTM helps to reduce the prediction duration with higher accuracy. It also enables the central system of a certain robotic application to prevent collisions caused by impediments in the interior or outdoor situation. The simulation results utilizing publicly available research datasets demonstrate the proposed model's efficiency and robustness compared to state-of-the-art approaches. The overall accuracy of the proposed model has improved approximately by 5% and reduced computational complexity by 84% approximately. [ABSTRACT FROM AUTHOR]

In this paper, we propose a novel video-sequence-based pedestrian re-identification method using the feature distribution similarity measurement between pedestrian video sequences (PRI-FDSM). We use the multiple granularity network combined with generative adversarial skew correction to extract and generate the feature point sets of the corrected pedestrian sequences. Then, we construct the corresponding probability function estimators for each pedestrian sequence using a radial basis function neural network to describe the feature distributions of specific sequences. Finally, we measure the similarity between the feature distributions of sequences to obtain re-identification results. The proposed method uses the distribution similarity measurement of the feature point sets of different sequences to make full use of all the image information of the specific pedestrian in a sequence. Thus, our method can mitigate the problem of insufficient use of the details of some images in a sequence, which commonly occurs in existing fusion feature point measurement methods. Besides, we correct the input skewed pedestrian sequences and achieve posture unification for the input sequences, which effectively mitigates the posture skewing problem of the photographed pedestrians in real-world surveillance scenes. We also build a dataset that more accurately represents the real-world surveillance scenes that contain pedestrian sequences with skewed postures. The results of the ablation experiment on iLIDS-VID and this dataset demonstrate the effectiveness of the proposed distribution-based similarity measurement method. We also compare the performance of the proposed method and several state-of-the-art methods on our dataset. Experimental results show that the indices of our method are all higher than those of the existing methods, and its mAP, Rank-1 and Rank-5 surpass the second best by 3.7%, 1.3% and 1.7% respectively. [ABSTRACT FROM AUTHOR]

Human pose estimation from video sequences has become a hot research topic in the domain of robotics and computer vision. However, existing three-dimensional (3D) pose estimation methods usually analyze individual frames, which have a low accuracy due to various human movement speed, limiting its practical application. In this paper, we propose a method for estimating 3D pose and calculating similarity from Tai Chi video sequences based on Seq2Seq network. Specifically, using 2D joint point coordinate sequence of the original image as input, our method constructs an encoder and a decoder to build a Seq2Seq network. Our method introduces an attention mechanism for weighing the input data to obtain an intermediate vector and decode it to estimate the 3D joint point sequence. Afterwards, using a template video and a target video as input, our method calculates the cost of passing through each point within the constraints to construct a cost matrix for video similarity. With the cost matrix, our method can determine the optimal path and use the correspondence of the video sequence to calculate the image similarity of the corresponding frame. The experimental data show that the proposed method can effectively improve the accuracy of 3D pose estimation, and increase the speed for video similarity calculation. [ABSTRACT FROM AUTHOR]

Outdoor images and videos suffer from several problems, such as the hazing problem due to the particles of dust, smoke, and other particles in the atmosphere. Videos in such atmospheric conditions are subject to visible quality degradations, such as low contrast and information loss. This paper presents a dehazing algorithm that enhances the video contrast, removes haze from hazy frames, and reduces frame degradation. We use a recursive deep residual learning (DRL) network as a dehazing tool to enhance video quality. The DRL network estimates the nonlinear mapping from the space of hazy input frames to that of output dehazed frames without estimating the transmission map and the atmospheric light as in traditional dehazing methods. After that, the dehazed frame is fed back to the input of the DRL network. This process is counted as an iteration. Our proposed algorithm depends on pre-processing of frames before the dehazing process to remove noise or enhance the visual quality, because all frames contain some noise due to sensor measurement errors. Noise can be amplified in the haze removal process if ignored. We use different types of enhancement techniques before the dehazing process. In addition, we modify the DRL network to be suitable for both near infrared (NIR) and visible frames. The number of iterations in the DRL network is increased from three iterations to nine iterations and the effect of increasing the number of iterations on the output dehazed frames is studied. We stopped at nine iterations, because elapsed time increases with the increase in the number of iterations. The peak signal-to-noise ratio and correlation after the dehazing process between dehazed and input hazy frames are used as evaluation metrics for our proposed algorithm. Results show that by increasing the number of iterations in the DRL network, dehazed frames record the best contrast, the highest spectral entropy, and the highest visual quality. [ABSTRACT FROM AUTHOR]

The fast progress in digital technology has sparked the generation of the amount of voluminous data from different social media platforms like Instagram, Facebook, YouTube, etc. There are other platforms, as well which generate large data like News, CCTV videos, sports, entertainment, etc. Lengthy Videos typically contain a significant number of duplicate occurrences that are uninteresting to the viewer. Eliminating this unnecessary information and concentra only on the crucial events will be far more advantageous. This produces a summary of lengthy films, which can save viewers time and enable better memory management. The highlights of a lengthy video are condensed into a video summary. Video summarization is an essential topic today, since many industries have CCTV cameras installed for various reasons such as monitoring, security, and tracking. Because surveillance videos are taken 24 hours a day, enormous amounts of memory and time are required if one wishes to trace any incident or person from the full day’s video. The summary generated from multiple views is far more challenging, so more study and advancement in MVS is required. The conceptual basis of video summarizing approaches is thoroughly addressed in this paper. This paper addresses applications and technology challenges in Single view and Multi View summarization.

The task of multi-object tracking via deep learning methods for UAV videos has become an important research direction. However, with some current multiple object tracking methods, the relationship between object detection and tracking is not well handled, and decisions on how to make good use of temporal information can affect tracking performance as well. To improve the performance of multi-object tracking, this paper proposes an improved multiple object tracking model based on FairMOT. The proposed model contains a structure to separate the detection and ReID heads to decrease the influence between every function head. Additionally, we develop a temporal embedding structure to strengthen the representational ability of the model. By combing the temporal-association structure and separating different function heads, the model's performance in object detection and tracking tasks is improved, which has been verified on the VisDrone2019 dataset. Compared with the original method, the proposed model improves MOTA by 4.9% and MOTP by 1.2% and has better tracking performance than the models such as SORT and HDHNet on the UAV video dataset. [ABSTRACT FROM AUTHOR]

The article analyzes the possible methods for performing laboratory work in chemistry during a situation of lockdown or pandemic. Laboratory practice is one of the main ways to learn the basics of science. The results of the experiments performed by the students stimulate their thinking, lead them to discussions and lead them to draw conclusions. All this leads to the optimization of the learning process. The use of very simple tools often helps in learning, facilitating the assimilation of the material. If the transfer of knowledge is not done correctly initially, then the subsequent understanding of the material will be more difficult to achieve. In the absence of a strong connection with previously studied topics it will be more difficult for students to think. Students begin to believe that science is difficult to understand, interest is often lost, and knowledge gained during the theoretical course may lose its value. People are naturally curious and always on the lookout, and the lab is a good place to satisfy that curiosity. In the laboratory, students have the opportunity to test their theoretical knowledge, which structures thinking and increases interest in science.

Motion estimation is considered significant for tracking the movement of an object in video sequences, and it is widely used in various video processing applications. Traditionally, many researchers focus on pixel-based motion estimation for object tracking, but it experienced increased computation time and cost. To reduce computation time, the utilization of a block-based motion estimation approach for object tracking is a recent trend. The existing block-based approach faces difficulty in finding representative points within the intensity domain. Therefore, this current research merged the deep learning approach with a block-matching algorithm for achieving efficient object tracking. In this proposed work, initially, video sequences are collected from a benchmark video dataset. Then, the acquired video sequences are segmented into frames. From the segmented frames, current and previous frames are considered for motion estimation. Frames are sent for the data augmentation process in which the process of flipping, cropping, and rotation is carried out. Then, the augmented frames are sent into Convolutional Neural Network (CNN) for feature extraction. Representative Point Matching (RPM) is used to estimate the motion vector based on the extracted features. After estimating the motion vector, the similarity between two consecutive frames is found using Structural Similarity Index (SSIM) technique. Finally, based on the similarity score, the movement of an object in the video is tracked effectively. Simulation analysis of the proposed block-based motion estimation model is done by evaluating some performance metrics. RMSE, PSNR, Execution Time, SSIM, and accuracy obtained for the proposed model are 27.5, 26.5 db , 31 sec , 0.91, and 94 %. This analysis suggested that the proposed CNN-RPM motion estimation model performs better in tracking the movement of the object. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Shanghai University of Engineering Science / Shanghai Gongcheng Jishu Daxue Xuebao is the property of Journal of Shanghai University of Engineering Science Editorial Office and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

The development of digital technology is utilized by people to capture and share video frames. At present, rather than capturing images, people are interested in recording video footage for exploring information. Here, retrieval of video from large databases is challenging due to the continuous frame count. To overcome these challenges associated with the retrieval of video from available databases, this research proposed a likelihood-based regression approach for video processing. To improve the retrieval accuracy of video sequences, the proposed method utilizes a likelihood estimation technique integrated with a regression model. The likelihood estimate measures the pixel level roughly for estimating the pixel range, after which the regression approach measures the pixel level for transforming certainly blurred and unwanted pixels. In the proposed likelihood regression approach, the video is converted into a video frame and stored in a database. Query frames are taken into account by the generated database depending on the features which are used for a given video to be retrieved. The significant video retrieval performance obtained from the simulation results for the proposed likelihood-based regression model shows that the proposed model performs well over the other state-of-the-art techniques. [ABSTRACT FROM AUTHOR]

This study tackles the semi‐supervised segmentation task for the objects that have large motion or appearance change in a video sequence, which is very challenging to the existing methods of video object segmentation (VOS). In this study, a novel adaptive approach is presented, named adaptive convolutional neural network for large change VOS, which determines when and how to fine‐tune the convolutional neural network through the motion metric and the appearance metric among consecutive video frames. Additionally, a lightweight optimisation algorithm for the predictive binary mask is introduced which is effective for pixel prediction by eliminating the discrete points cluster. To illustrate the advantages of this approach, experiments have been performed on four VOS datasets, which demonstrate that the proposed method is highly effective and could achieve the state‐of‐the‐art on these datasets.

Existing person re‐identification (re‐id) models mainly focus on still‐image‐based module, namely matching person images across non‐overlapping camera views. Since video sequence contains much more information than still images and can be easily achieved by tracking algorithms in practical applications, the video re‐id has attracted increasing attention in recent years. Distance learning is crucial for a re‐id system. However, the computed distances in traditional video‐based methods are easily distracted by the randomness of data distribution, especially with small sample size for training. To preferably distinguish different people, a novel regularised hull distance learning video‐based person re‐id method is proposed. It is advantageous in two aspects: robustness is guaranteed due to expanded video samples by regularised affine hull with limited ones, discriminability is ensured due to penalised hard negative samples more severely. Hence, the discriminability and robustness of the learnt metric are strengthened. Comparisons with the state‐of‐the‐art video‐based methods as well as related methods on PRID 2011, iLIDS‐VID and MARS datasets demonstrate the superiority of the authors’ method.

Abstract With the advent of the era of big data, the Internet industry has produced massive amounts of multimedia video data. In order to process these video sequence data quickly and effectively, a visual information extraction method based on wavelet transform and feature comparison is proposed to perceive the target of interest by simulating the multi-channel spatial frequency decomposition function of human visual system which can quickly extract the significant distribution from the image and obtain region of interest (ROI). Firstly, the principle of visual attention mechanism and the visual salience detection were analyzed. Then, the DOG (Difference of Gaussian) function is taken as the wavelet basis function, and the wavelet data is used to decompose the image data in the spatial domain and frequency domain, thus applying to the multi-channel of the human visual system. Finally, the significant distribution in the entire image is obtained by global color comparison; thus, the region of interest is extracted. The extraction model of visual information proposed is simulated in MATLAB environment. The simulation results show that the proposed algorithm can extract ROI more accurately and efficiently, compared with the existing algorithms.

Sign language recognition aims to recognize meaningful movements of hand gestures and is a significant solution in intelligent communication between the deaf community and hearing societies. However, until now, the current dynamic sign language recognition methods still have some drawbacks with difficulties of recognizing complex hand gestures, low recognition accuracy for most dynamic sign language recognition, and potential problems in larger video sequence data training. In order to solve these issues, this paper presents a multimodal dynamic sign language recognition method based on a deep 3-dimensional residual ConvNet and bi-directional LSTM networks, which is named as BLSTM-3D residual network (B3D ResNet). This method consists of three main parts. First, the hand object is localized in the video frames in order to reduce the time and space complexity of network calculation. Then, the B3D ResNet automatically extracts the spatiotemporal features from the video sequences and establishes an intermediate score corresponding to each action in the video sequence after feature analysis. Finally, by classifying the video sequences, the dynamic sign language is accurately identified. The experiment is conducted on test datasets, including DEVISIGN_D dataset and SLR_Dataset. The results show that the proposed method can obtain state-of-the-art recognition accuracy (89.8% on the DEVISIGN_D dataset and 86.9% on SLR_Dataset). In addition, the B3D ResNet can effectively recognize complex hand gestures through larger video sequence data, and obtain high recognition accuracy for 500 vocabularies from Chinese hand sign language.

The stability and quantity of feature matching in video sequence is one of the key issues for feature tracking and some relevant applications. The existing matching methods are based on feature detection, which is usually affected by illumination conditions, noise or occlusions, and this will directly influence matching results. In this paper, we propose an accurate prediction method for interest point estimation in video sequence by extracting the stable mapping for each undetected point in its suitable projective plane, which is based on coplanar feature points that have already been detected in adjacent frames. The proposed prediction method breaks the limitation of the previous approaches that largely rely on feature detection. Our experiments show that our method not only predicts features accurately, but also enriches the correspondences, which prolongs the track length of features. [ABSTRACT FROM AUTHOR]

Moving object detection in a video sequence is one of the leading tasks of marine scientists to explore and monitor applications. The videos acquired in the underwater environment are usually degraded due to the physical properties of water medium as compared with images acquired in the air and that affects the performance of feature descriptors. In this study, a new feature descriptor, multi‐frame triplet pattern (MFTP) is proposed for underwater moving object detection. The MFTP encodes the structure of local region based on three sets of frames, which are calculated by considering local differences in intensities between the centre pixel and its nine neighbours. Furthermore, the robustness of the proposed method is increased by integrating it with colour and motion features. The performance of the proposed framework is tested by conducting seven experiments on Fish4Knowledge database for underwater moving object detection applications. The results of the proposed method show a significant improvement as compared with state‐of‐the‐art techniques in terms of their evaluation measures.

Gesture is a natural form of human communication, and it is of great significance in human–computer interaction. In the dynamic gesture recognition method based on deep learning, the key is to obtain comprehensive gesture feature information. Aiming at the problem of inadequate extraction of spatiotemporal features or loss of feature information in current dynamic gesture recognition, a new gesture recognition architecture is proposed, which combines feature fusion network with variant convolutional long short‐term memory (ConvLSTM). The architecture extracts spatiotemporal feature information from local, global and deep aspects, and combines feature fusion to alleviate the loss of feature information. Firstly, local spatiotemporal feature information is extracted from video sequence by 3D residual network based on channel feature fusion. Then the authors use the variant ConvLSTM to learn the global spatiotemporal information of dynamic gesture, and introduce the attention mechanism to change the gate structure of ConvLSTM. Finally, a multi‐feature fusion depthwise separable network is used to learn higher‐level features including depth feature information. The proposed approach obtains very competitive performance on the Jester dataset with the classification accuracies of 95.59%, achieving state‐of‐the‐art performance with 99.65% accuracy on the SKIG (Sheffifield Kinect Gesture) dataset. [ABSTRACT FROM AUTHOR]