Search

Visual domain adaptation can utilize similar but different source data to solve the scarcity problem of labeled data in target domain, therefore it has received more and more attention. The main difficulty of this task is the distribution discrepancy between target data and source data. Hence, a series of works have been proposed to reduce the distribution discrepancy by seeking cross-domain feature invariant, or reducing the weight of source instances that are not related to the target data. The major disadvantage is that most works fail to consider the intrinsic geometric manifold structure, which is important for effective learning. This paper considers the intrinsic geometric manifold structure, and the obtained domain invariant features can be more discriminating. Our method effectively integrates distribution matching and preserving the structure consistency into one framework, and obtains the optimal solution by solving a generalized eigen-decomposition problem. Comprehensive experimental results verify that proposed method can significantly outperform state-of-the-art methods in cross domain image recognition problems.

Object recognition finds wide application and is a challenge topic in computer vision and pattern recognition. In this paper, a new object detection method is proposed, which depends on active contour models (ACMs) energy function with pattern matching method. Firstly, object detection with shape prior based on energy function is presented to extract candidate targets. A novel geometric feature based on energy function of prior shape is put forward. The object detection method is invariant to scaling of shape and is also suitable for prior shapes set for general cases. Several candidate targets may be extracted when the object detection method is applied in object detection because most objects have similar shape. Then Speed up Robust Features (SURF) algorithm is employed to recognize the desired target from candidate targets. Experimental results show that proposed method is effective and efficient.

Organic light-emitting diode(OLED) has emerged as the new generation technique to mobile displays. Although OLED plane has many advantages compared to LCD, it still accounts for a large amount of the power consumption in mobile devices. In this paper, we propose a log-function power constrained algorithm(LFPC) to save the power of OLED display. We choose superpixel as the basic unit and computer a saliency map of all superpixels. Next, we integrate the human visual characteristic into a log-function which is used to limit the pixel values thus achieving the power saving of OLED. Our experimental results show that LFPC algorithm can achieve 34.6% more OLED power saving for images with high luminance while retaining a high image visual experience. Also, our methodology performs better on power saving and image quality compared to PCCE.

This article is about vehicles detection from visual data which is an important part of automotive driving assistance systems. It is a big challenge to make the vehicles detection more robust. In order to enhance the robustness, a lane lines stabilization methods is proposed by studying the imaging model. Next, a hypothesis generation and verification framework is applied for saving time. Finally, our approach is compared to the Entropy-verified method and DPM in runtime performance and false positive ratio. It turns out that our approach gets a balance between runtime performance and false positive ratio.