Limits...
Effective Visual Tracking Using Multi-Block and Scale Space Based on Kernelized Correlation Filters

View Article: PubMed Central - PubMed

ABSTRACT

Accurate scale estimation and occlusion handling is a challenging problem in visual tracking. Recently, correlation filter-based trackers have shown impressive results in terms of accuracy, robustness, and speed. However, the model is not robust to scale variation and occlusion. In this paper, we address the problems associated with scale variation and occlusion by employing a scale space filter and multi-block scheme based on a kernelized correlation filter (KCF) tracker. Furthermore, we develop a more robust algorithm using an appearance update model that approximates the change of state of occlusion and deformation. In particular, an adaptive update scheme is presented to make each process robust. The experimental results demonstrate that the proposed method outperformed 29 state-of-the-art trackers on 100 challenging sequences. Specifically, the results obtained with the proposed scheme were improved by 8% and 18% compared to those of the KCF tracker for 49 occlusion and 64 scale variation sequences, respectively. Therefore, the proposed tracker can be a robust and useful tool for object tracking when occlusion and scale variation are involved.

No MeSH data available.


Related in: MedlinePlus

Tracking results for partial occlusions for the sequences (a) Walking2; (b) FaceOcc1; (c) Human6; (d) Girl; (e) Subway; and (f) Rubik.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC5375719&req=5

sensors-17-00433-f008: Tracking results for partial occlusions for the sequences (a) Walking2; (b) FaceOcc1; (c) Human6; (d) Girl; (e) Subway; and (f) Rubik.

Mentions: The factors of occlusion, scale variation and above these illumination variation, deformation, and fast motion, affect the performance of the tracking algorithms. Scale variation implies a change in the target size. In Figure 7, the images Singer1, Dog1, and Human4, are typical sequences with the scale variation attribute. However, in Figure 8, the Walking2 and Human6 sequences have scale variation and partial occlusion at the same time. Thus, each of the tracking attributes exists in a complex manner. Among the attributes, occlusion occurs frequently in tracking. Heavy occlusion implies that the object is covered in its entirety; therefore, it is difficult to control with tracking. On the other hand, partial occlusion occurs when regions of the object remain visible, and therefore, in this case tracking remains possible. In Figure 8, the target in the video FaccOcc1 is partially occluded. In the Walking2 sequence, the target is covered by a walking man, but approximately one-third of the target object remains visible. Regions such as this that remain partially visible throughout a sequence of images are considered reliable regions and are selected by the proposed multi-block model. Thus, the tracking result for the Walking2 sequence was successful, whereas in the Struck and VTD sequences, the tracking algorithm loses the woman at times during which she is occluded by the man, but approximately one-third of the woman remains visible. Human3, Human4, and Human6 are outdoor sequences. These outdoor images are frequently affected by partial occlusion, scale variation, and fast motion. In Figure 7 and Figure 8, the results show that the tracking procedure of the proposed method is more successful than any other method. Figure 9 presents a comparison of the most successful state-of-the-art trackers. Each sequence includes plural attributes. This resulted in degraded performance, even though the method is robust against occlusion. The proposed algorithm is able to overcome occlusion and scale variation, and outperforms other trackers.


Effective Visual Tracking Using Multi-Block and Scale Space Based on Kernelized Correlation Filters
Tracking results for partial occlusions for the sequences (a) Walking2; (b) FaceOcc1; (c) Human6; (d) Girl; (e) Subway; and (f) Rubik.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC5375719&req=5

sensors-17-00433-f008: Tracking results for partial occlusions for the sequences (a) Walking2; (b) FaceOcc1; (c) Human6; (d) Girl; (e) Subway; and (f) Rubik.
Mentions: The factors of occlusion, scale variation and above these illumination variation, deformation, and fast motion, affect the performance of the tracking algorithms. Scale variation implies a change in the target size. In Figure 7, the images Singer1, Dog1, and Human4, are typical sequences with the scale variation attribute. However, in Figure 8, the Walking2 and Human6 sequences have scale variation and partial occlusion at the same time. Thus, each of the tracking attributes exists in a complex manner. Among the attributes, occlusion occurs frequently in tracking. Heavy occlusion implies that the object is covered in its entirety; therefore, it is difficult to control with tracking. On the other hand, partial occlusion occurs when regions of the object remain visible, and therefore, in this case tracking remains possible. In Figure 8, the target in the video FaccOcc1 is partially occluded. In the Walking2 sequence, the target is covered by a walking man, but approximately one-third of the target object remains visible. Regions such as this that remain partially visible throughout a sequence of images are considered reliable regions and are selected by the proposed multi-block model. Thus, the tracking result for the Walking2 sequence was successful, whereas in the Struck and VTD sequences, the tracking algorithm loses the woman at times during which she is occluded by the man, but approximately one-third of the woman remains visible. Human3, Human4, and Human6 are outdoor sequences. These outdoor images are frequently affected by partial occlusion, scale variation, and fast motion. In Figure 7 and Figure 8, the results show that the tracking procedure of the proposed method is more successful than any other method. Figure 9 presents a comparison of the most successful state-of-the-art trackers. Each sequence includes plural attributes. This resulted in degraded performance, even though the method is robust against occlusion. The proposed algorithm is able to overcome occlusion and scale variation, and outperforms other trackers.

View Article: PubMed Central - PubMed

ABSTRACT

Accurate scale estimation and occlusion handling is a challenging problem in visual tracking. Recently, correlation filter-based trackers have shown impressive results in terms of accuracy, robustness, and speed. However, the model is not robust to scale variation and occlusion. In this paper, we address the problems associated with scale variation and occlusion by employing a scale space filter and multi-block scheme based on a kernelized correlation filter (KCF) tracker. Furthermore, we develop a more robust algorithm using an appearance update model that approximates the change of state of occlusion and deformation. In particular, an adaptive update scheme is presented to make each process robust. The experimental results demonstrate that the proposed method outperformed 29 state-of-the-art trackers on 100 challenging sequences. Specifically, the results obtained with the proposed scheme were improved by 8% and 18% compared to those of the KCF tracker for 49 occlusion and 64 scale variation sequences, respectively. Therefore, the proposed tracker can be a robust and useful tool for object tracking when occlusion and scale variation are involved.

No MeSH data available.


Related in: MedlinePlus