Entity

Time filter

Source Type


Zhao C.,CAS Shenyang Institute of Automation | Zhao C.,Chinese Academy of Sciences | Zhao C.,The Key Laboratory of Image Understanding and Computer Vision | Zhao C.,University of Chinese Academy of Sciences | And 3 more authors.
Jisuanji Fuzhu Sheji Yu Tuxingxue Xuebao/Journal of Computer-Aided Design and Computer Graphics | Year: 2015

SURF algorithm has high computational complexity, and requires a lot of logic and memory resources. Moreover the process of descriptor extraction is difficult to implement in parallel and unable to meet real-time requirements. To solve the above disadvantages, an optimized SURF algorithm is put forward and the FPGA implementation is also provided. A rotation invariant and fully parallel optimized SURF algorithm is achieved using circular feature region and radial gradient transform method, which cancels the processes of main direction calculation and feature region rotation. Then the optimized SURF algorithm is implemented based on FPGA by using multi-memory and multi-channel parallel pipelined architecture. By experimental comparison, the matching performance of the optimized SURF algorithm is as good as the original SURF algorithm. Compared with the original SURF descriptor, the number of matching points reduces in 5% to 20%, but the accuracy of matching improves in 5% to 10%. The FPGA implementation of proposed SURF algorithm meets real-time requirements by using 13.5 MHz clock. For a video stream with resolution of 720×576, the processing speed reaches 25 fps. ©, 2015, Institute of Computing Technology. All right reserved. Source


Shao C.,CAS Shenyang Institute of Automation | Shao C.,University of Chinese Academy of Sciences | Shao C.,Chinese Academy of Sciences | Ding Q.,Space Star Technology | And 4 more authors.
Hongwai yu Jiguang Gongcheng/Infrared and Laser Engineering | Year: 2016

Inspired by the fact that a rigid body has consistent transformation for its individual part, a novel target tracking algorithm based on high-dimension data clustering is proposed. The proposed measure is proved to be available in object tracking mathematically. Thus, it is called the High- Dimension Data Clustering (HDDC) tracker. The frameworks of proposed method are as follows. First, Harris detector is utilized to extract the corners both in the template and the tracking region. Second, these feature points are grouped via their position information separately. Third, affine matrixes between the template and the tracking region are calculated among their respective feature groups. At last, high-dimension data clustering is carried out to measure these matrixes, and the feature points corresponding with the similar matrixes that are tracked targets. Extensive experimental results demonstrate that HDDC is efficient on measuring affine deformed objects and outperforms some state-of-the-art discriminative tracking methods. © 2016, Editorial Board of Journal of Infrared and Laser Engineering. All right reserved. Source


Sun Z.,CAS Shenyang Institute of Automation | Sun Z.,University of Chinese Academy of Sciences | Sun Z.,Chinese Academy of Sciences | Sun Z.,The Key Laboratory of Image Understanding and Computer Vision | And 13 more authors.
Hongwai yu Jiguang Gongcheng/Infrared and Laser Engineering | Year: 2015

A simple and computationally efficient method was presented for detecting visually salient objects in infrared radiation images. The proposed method can be divided into three steps. Firstly, the infrared image was pre-processed to increase the contrast between objects and background. Secondly, the spectral residual of the pre-processed image was extracted in the log spectrum, then via corresponding inverse transform and threshold segmentation we could get the rough regions of the salient objects. Finally, a sliding window was applied to acquire the explicit position of the salient objects using the probabilistic interpretation of the semi-local feature contrast which was estimated by comparing the gray level distribution of the object and the surrounding area in the original image. And changing the size of the sliding window, different size of objects could be found out. The method was tested on abundant amount of infrared radiation images, and the results show that the saliency detection based object detection method is effective and robust. ©, 2015, Editorial Board of Journal of Infrared and Laser Engineering. All right reserved. Source


Zhao C.,CAS Shenyang Institute of Automation | Zhao C.,Chinese Academy of Sciences | Zhao C.,The Key Laboratory of Image Understanding and Computer Vision | Zhao C.,University of Chinese Academy of Sciences | And 15 more authors.
Neurocomputing | Year: 2016

Although a number of local feature-based methods have been proposed, the multimodality matching is still a challenging problem in object recognition, remote sensing and medical image processing where the image contrast is significantly different. The local feature-based multimodality matching method is usually intensity-based, so the matching performance is not good enough because intensity-based method is sensitive to contrast variations. In order to solve these problems, we propose a novel Multimodality Robust Line Segment Descriptor (MRLSD) and develop a MRLSD matching method. The proposed method generates MRLSD descriptors based on extracted highly equivalent corners and line segments for two multimodal images, and then performs image matching by measuring the similarity of corresponding descriptors over two images. The proposed corner and line segment extraction method is based on local phase and direction information, and is insensitive to contrast variations, so the MRLSD descriptor is robust to modality variations. The MRLSD descriptor is rotation invariant by selecting circular feature sub-regions and projecting feature vectors to radial direction. The MRLSD descriptor achieves scale invariance by adjusting the radius of circular feature region according to the scale. Experimental results indicate that the proposed method achieves higher precision and repeatability than several state-of-the-art local feature-based multimodality matching methods, and also demonstrate its robustness to multimodal images. © 2015 Elsevier B.V. Source

Discover hidden collaborations