Directional Prediction CamShift algorithm based on Adaptive Search Pattern for moving object tracking

Hsia, Chih-Hsien; Liou, Yun-Jung; Chiang, Jen‐Shiun

doi:10.1007/s11554-013-0382-x

Cited by 20 publications

(11 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The second part is a random process, and the matrix B represents the probability distribution description of the state. Hidden Markov models can be represented by quintuple [50]- [52]: A-The time-independent state transition probability matrix is:A = a ij , a ij = P q t+1 = s j q t = s i , 1 ≤ i, j ≤ N , and N j=1 a ij = 1; B-state output probability matrix is:…”

Section: Model Trainingmentioning

confidence: 99%

Dynamic Gesture Recognition in the Internet of Things

Jiang

et al. 2019

IEEE Access

View full text Add to dashboard Cite

Gesture recognition based on computer vision has gradually become a hot research direction in the field of human-computer interaction. The field of human-computer interaction is an important direction in the Internet of Things (IoTs) technology. Human-computer interaction through gestures is the direction of continuous research on IoTs technology. In recent years, the Kinect sensor-based gesture recognition method has been widely used in gesture recognition, because it can separate gestures from complex backgrounds and is less affected by illumination and can accurately track and locate gesture motions. At present, the Kinect sensor needs to be further improved on the recognition of complex gesture movements, especially the problem that the recognition rate of dynamic gestures is not high, which hinders the development of humancomputer interaction under the IoTs technology. In this paper, based on the above problems, the Kinectbased gesture recognition is analyzed in detail, and a dynamic gesture recognition method based on HMM and D-S evidence theory is proposed. Based on the original HMM, the tangent angle and gesture change at different moments of the palm trajectory are used as the characteristics of the complex motion gesture, and the dimension of the trajectory tangent is reduced by the number of quantization codes. Then, the parameter model training of HMM is completed. Finally, combined with D-S evidence theory, combinatorial logic is judged, dynamic gesture recognition is carried out, and a better recognition effect is obtained, which lays a good foundation for human-computer interaction under the IoTs technology. INDEX TERMSGesture recognition, hidden Markov model (HMM), D-S evidence theory, Internet of Things (IoT).

show abstract

Section: Model Trainingmentioning

confidence: 99%

Dynamic Gesture Recognition in the Internet of Things

Jiang

et al. 2019

IEEE Access

View full text Add to dashboard Cite

show abstract

“…For each captured frame, the original image is converted to a probabilistic color distribution of the image, which using a histogram model of the color to be traced. 12 In this case, the characteristic color to be tracked is that of the hand, after having recognized this pattern, the desired location is made in the scene, and then the segmentation of this as shown in Four gestures were used to control the robots Figure 4.a, each gesture transmits an action to the robots being able to be of two categories, collective or individual. The gesture "G1" conveys a collective action for the robots which consists of the same ones executing a forward movement movement, "G2" just like the previous gesture transmits a collective action causing the robots to return to the initial position, "G3" by its Instead it consists of an individual action that causes the robot from the lower right corner to move to the center, "G4" transmits an individual action that causes the robot from the upper left corner to move up to the scepter.…”

Section: Methodsmentioning

confidence: 99%

Navigation of quadruped multi-robots by gesture recognition using restricted Boltzmann machines

Saraiva¹,

Santos

Marqués

et al. 2018

Robotics Transforming the Future

View full text Add to dashboard Cite

show abstract

“…An improved Camshift algorithm based on Kalman filtering is presented to realise robust and fast video tracking [2]. A directional prediction Camshift algorithm that uses an adaptive search pattern is proposed in [3] to refine the central area search and improve the tracking accuracy rate. An improved Camshift which adopts pixel filtering, histogram improvement, occlusion judgement, and Kalman filtering to enhance the tracking accuracy and reliability is presented in [4].…”

Section: Introductionmentioning

confidence: 99%

“…All of the above‐mentioned algorithms cannot automatically select the initial locations of video targets. The adaptive kernel size and searching window size are not supported in [1–4]. The tracking accuracy in poor illumination cannot be guaranteed in [2–4, 6, 7].…”

Section: Introductionmentioning

confidence: 99%

Robust video tracking algorithm: a multi‐feature fusion approach

Wang

Nguang

Wen

2018

IET Computer Vision

View full text Add to dashboard Cite

This study proposes a novel robust video tracking algorithm consists of target detection, multi-feature fusion, and extended Camshift. Firstly, a novel target detection method that integrates Canny edge operator, three-frame difference, and improved Gaussian mixture model (IGMM)-based background modelling is provided to detect targets. The IGMM-based background modelling divides video frames into meshes to avoid pixel-wise processing. In addition, the output of the target detection is utilised to initialise the IGMM and to accelerate the convergence of iterations. Secondly, low-dimensional regional covariance matrices are introduced to describe video targets by fusing multiple features like pixel location, colour index, rotation and scale invariant features as well as uniform local binary patterns, and directional derivatives. Thirdly, an extended Camshift based on adaptive kernel bandwidth and robust H ∞ state estimation is proposed to predict the states of fast moving targets and to reduce the mean shift iterations. Finally, the effectiveness of the proposed tracking algorithm is demonstrated via experiments. 2 Automatic detection of video targets Researchers have proposed various target detection methods such as background subtraction (BGS), adjacent frames difference, and target modelling based on appearance. In this section, an automatic target detection method both takes care of the efficiency and the accuracy is presented as shown in Fig. 1, where I k and B k , respectively, represent the image frame and the background at

show abstract

Directional Prediction CamShift algorithm based on Adaptive Search Pattern for moving object tracking

Cited by 20 publications

References 17 publications

Dynamic Gesture Recognition in the Internet of Things

Dynamic Gesture Recognition in the Internet of Things

Navigation of quadruped multi-robots by gesture recognition using restricted Boltzmann machines

Robust video tracking algorithm: a multi‐feature fusion approach

Contact Info

Product

Resources

About