A novel ELM based adaptive Kalman filter tracking algorithm

Ren

2018

Multimed Tools Appl

The tracking method based on the extreme learning machine (ELM) is efficient and effective. ELM randomly generates input weights and biases in the hidden layer, and then calculates and computes the output weights by reducing the iterative solution to the problem of linear equations. Therefore, ELM offers the satisfying classification performance and fast training time than other discriminative models in tracking. However, the original ELM method often suffers from the problem of the imbalanced classification distribution, which is caused by few target objects, leading to under-fitting and more background samples leading to over-fitting. Worse still, it reduces the robustness of tracking under special conditions including occlusion, illumination, etc. To address above problems, in this paper, we present a robust tracking algorithm. First, we introduce the local weight matrix that is the dynamic creation from the data distribution at the current frame in the original ELM so as to balance between the empirical and structure risk, and fully learn the target object to enhance the classification performance. Second, we improve it to the incremental learning method ensuring tracking real-time and efficient. Finally, the forgetting factor is used to strengthen the robustness for changing of the classification distribution with time. Meanwhile, we propose a novel optimized method to obtain the optimal sample as the target object, which avoids track-Jing Zhang School of computer and information technology,

“…According to above expression (13), and A T N +∆N B N +∆N = A T N B N +A T ∆N B ∆N . We can get as the following expression:…”

Section: And Then the Equation Can Change Intoβ = (mentioning

confidence: 94%

Section: Introductionmentioning

confidence: 99%

Robust tracking via weighted online extreme learning machine

Zhang

Ren

2018

Multimed Tools Appl

“…Although this method is effective and easy to implement, the cost of the system setup and maintenance is high. Chi et al [ 34 ] used an extreme learning machine (ELM) to improve the estimation accuracy and the robustness of traditional KF for tracking in WSNs. Mahfouz et al [ 20 ] trained the ridge regression (RR) and the vector-output regularized least-squares algorithms off-line to obtain a first estimation position.…”

Section: Related Workmentioning

confidence: 99%

SVM+KF Target Tracking Strategy Using the Signal Strength in Wireless Sensor Networks

Liu

et al. 2020

Sensors

Target Tracking (TT) is a fundamental application of wireless sensor networks. TT based on received signal strength indication (RSSI) is by far the cheapest and simplest approach, but suffers from a low stability and precision owing to multiple paths, occlusions, and decalibration effects. To address this problem, we propose an innovative TT algorithm, known as the SVM+KF method, which combines the support vector machine (SVM) and an improved Kalman filter (KF). We first use the SVM to obtain an initial estimate of the target’s position based on the RSSI. This enhances the ability of our algorithm to process nonlinear data. We then apply an improved KF to modify this estimated position. Our improved KF adds the threshold value of the innovation update in the traditional KF. This value changes dynamically according to the target speed and network parameters to ensure the stability of the results. Simulations and real experiments in different scenarios demonstrate that our algorithm provides a superior tracking accuracy and stability compared to similar algorithms.

Proceedings of the 2016 4th International Conference on Machinery, Materials and Computing Technology

“…KSTC.As known in the world, Kalman Filter is used to calculate the optimal value of next state by combining the estimated and measurement [6] . As STC can track object effectively without occlusion, take the output of it as Kalman Filter's measurement [7] .…”

Section: Stcmentioning

confidence: 99%

STC Tracking Algorithm Based on Kalman Filter

Chen¹,

Yang²

2016

Abstract. During object tracking, Fast tracking via Spatio-Temporal Context Learning which combines temporal correlation among sequential frames and spatial correlation between object and background can solve the problem of semi-occlusion, but not full-occlusion. Kalman Filter makes use of the predictive value and measurement to calculate the optimal state. This paper aims at solving the full-occlusion problems by combining the algorithm and Kalman Filter together. Experiments show that the improved STC can solve occlusion problems effectively.