Zheng Chang

Image interpolation for division of focal plane polarimeters with intensity correlation.

Division of focal plane (DoFP) polarimeters operate by integrating micro-polarizer elements with a focal plane. These polarization imaging sensors reduce spatial resolution output and each pixel has a varying instantaneous field of view (IFoV). These drawbacks can be mitigated by applying proper interpolation methods. In this paper, we present a new interpolation method for DoFP polarimeters by using intensity correlation. We employ the correlation of intensity measurements in different orientations to detect edges and then implement interpolation along edges. The performance of the proposed method is compared with several previous methods by using root mean square error (RMSE) comparison and visual comparison. Experimental results showed that our proposed method can achieve better visual effects and a lower RMSE than other methods.

Pedestrian detection with semantic regions of interest

For many pedestrian detectors, background vs. foreground errors heavily influence the detection quality. Our main contribution is to design semantic regions of interest that extract the foreground target roughly to reduce the background vs. foreground errors of detectors. First, we generate a pedestrian heat map from the input image with a full convolutional neural network trained on the Caltech Pedestrian Dataset. Next, semantic regions of interest are extracted from the heat map by morphological image processing. Finally, the semantic regions of interest divide the whole image into foreground and background to assist the decision-making of detectors. We test our approach on the Caltech Pedestrian Detection Benchmark. With the help of our semantic regions of interest, the effects of the detectors have varying degrees of improvement. The best one exceeds the state-of-the-art.

Real-Time Robust Tracking for Motion Blur and Fast Motion via Correlation Filters

Visual tracking has extensive applications in intelligent monitoring and guidance systems. Among state-of-the-art tracking algorithms, Correlation Filter methods perform favorably in robustness, accuracy and speed. However, it also has shortcomings when dealing with pervasive target scale variation, motion blur and fast motion. In this paper we proposed a new real-time robust scheme based on Kernelized Correlation Filter (KCF) to significantly improve performance on motion blur and fast motion. By fusing KCF and STC trackers, our algorithm also solve the estimation of scale variation in many scenarios. We theoretically analyze the problem for CFs towards motions and utilize the point sharpness function of the target patch to evaluate the motion state of target. Then we set up an efficient scheme to handle the motion and scale variation without much time consuming. Our algorithm preserves the properties of KCF besides the ability to handle special scenarios. In the end extensive experimental results on benchmark of VOT datasets show our algorithm performs advantageously competed with the top-rank trackers.

Non-uniformity correction for division of focal plane polarimeters with a calibration method.

Division of focal plane polarimeters are composed of nanometer polarization elements overlaid upon a focal plane array (FPA) sensor. The manufacturing flaws of the polarization grating and each detector in the FPA having a different photo response can introduce non-uniformity errors when reconstructing the polarization image without correction. A new calibration method is proposed to mitigate non-uniformity errors in the visible waveband. We correct non-uniformity in the form of a vector. The correction matrix and offset vector are calculated for the following correction. The performance of the proposed method is compared with state-of-the-art techniques by employing simulated data and real scenes. The experimental results showed that the proposed method can effectively mitigate non-uniformity errors and achieve better visual results.

Robust Scale Adaptive Tracking by Combining Correlation Filters with Sequential Monte Carlo

A robust and efficient object tracking algorithm is required in a variety of computer vision applications. Although various modern trackers have impressive performance, some challenges such as occlusion and target scale variation are still intractable, especially in the complex scenarios. This paper proposes a robust scale adaptive tracking algorithm to predict target scale by a sequential Monte Carlo method and determine the target location by the correlation filter simultaneously. By analyzing the response map of the target region, the completeness of the target can be measured by the peak-to-sidelobe rate (PSR), i.e., the lower the PSR, the more likely the target is being occluded. A strict template update strategy is designed to accommodate the appearance change and avoid template corruption. If the occlusion occurs, a retained scheme is allowed and the tracker refrains from drifting away. Additionally, the feature integration is incorporated to guarantee the robustness of the proposed approach. The experimental results show that our method outperforms other state-of-the-art trackers in terms of both the distance precision and overlap precision on the publicly available TB-50 dataset.

Robust visual tracking of infrared object via sparse representation model

In this paper, we propose a robust tracking method for infrared object. We introduce the appearance model and the sparse representation in the framework of particle filter to achieve this goal. Representing every candidate image patch as a linear combination of bases in the subspace which is spanned by the target templates is the mechanism behind this method. The natural property, that if the candidate image patch is the target so the coefficient vector must be sparse, can ensure our algorithm successfully. Firstly, the target must be indicated manually in the first frame of the video, then construct the dictionary using the appearance model of the target templates. Secondly, the candidate image patches are selected in following frames and the sparse coefficient vectors of them are calculated via ℓ1-norm minimization algorithm. According to the sparse coefficient vectors the right candidates is determined as the target. Finally, the target templates update dynamically to cope with appearance change in the tracking process. This paper also addresses the problem of scale changing and the rotation of the target occurring in tracking. Theoretic analysis and experimental results show that the proposed algorithm is effective and robust.

Sky Detection in Hazy Image

Sky detection plays an essential role in various computer vision applications. Most existing sky detection approaches, being trained on ideal dataset, may lose efficacy when facing unfavorable conditions like the effects of weather and lighting conditions. In this paper, a novel algorithm for sky detection in hazy images is proposed from the perspective of probing the density of haze. We address the problem by an image segmentation and a region-level classification. To characterize the sky of hazy scenes, we unprecedentedly introduce several haze-relevant features that reflect the perceptual hazy density and the scene depth. Based on these features, the sky is separated by two imbalance SVM classifiers and a similarity measurement. Moreover, a sky dataset (named HazySky) with 500 annotated hazy images is built for model training and performance evaluation. To evaluate the performance of our method, we conducted extensive experiments both on our HazySky dataset and the SkyFinder dataset. The results demonstrate that our method performs better on the detection accuracy than previous methods, not only under hazy scenes, but also under other weather conditions.

PCA-based denoising method for division of focal plane polarimeters

Division of focal plane (DoFP) polarimeters are composed of interlaced linear polarizers overlaid upon a focal plane array sensor. The interpolation is essential to reconstruct polarization information. However, current interpolation methods are based on the unrealistic assumption of noise-free images. Thus, it is advantageous to carry out denoising before interpolation. In this paper, we propose a principle component analysis (PCA) based denoising method, which works directly on DoFP images. Both simulated and real DoFP images are used to evaluate the denoising performance. Experimental results show that the proposed method can effectively suppress noise while preserving edges.

Discriminative feature selection for visual tracking

Visual tracking is an important role in computer vision tasks. The robustness of tracking algorithm is a challenge. Especially in complex scenarios such as clutter background, illumination variation and appearance changes etc. As an important component in tracking algorithm, the appropriateness of feature is closed related to the tracking precision. In this paper, an online discriminative feature selection is proposed to provide the tracker the most discriminative feature. Firstly, a feature pool which contains different information of the image such as gradient, gray value and edge is built. And when every frame is processed during tracking, all of these features will be extracted. Secondly, these features are ranked depend on their discrimination between target and background and the highest scored feature is chosen to represent the candidate image patch. Then, after obtaining the tracking result, the target model will be update to adapt the appearance variation. The experiment show that our method is robust when compared with other state-of-the-art algorithms.

Step-by-step pipeline processing approach for line segment detection

This study proposes a line segment detection that can efficiently and effectively handle non-linear uniform intensity changes. The presented sketching algorithm applies the resistant to affine transformation and monotonic intensity change (RATMIC) descriptor to conduct binary translation in the image pre-processing step, which can remove the unwanted smoothing of the Canny detector in most line detections. The Harris corner detector is applied to catch regions of line segments for the purpose of simulating the composition of sketching and achieving a sense of unity within the picture. Furthermore, the RATMIC descriptor is employed to obtain binary images of the regions of interest (ROIs). Finally, small eigenvalue analysis is implemented to detect straight lines in the ROIs. The experiments conducted on various images with image rotation, scaling, and translation validate the effectiveness of the proposed method. The experimental results also demonstrate that about 30% in the overall coverage of major lines and 20% in the coverage per major line are increased compared with the state-of-the-art line detectors. Moreover, the performance of the proposed method produces a combined advantage of approximate to 17% in the coverage of line segments over the line segment detector with noisy images.