• Title, Summary, Keyword: 고정 특징점 추출기

Search Result 4, Processing Time 0.027 seconds

The Background Modeling Method under Camera Shaking (카메라 흔들림을 고려한 배경 모델 생성 방법)

  • Lee, Jaehoon;Kim, Hyungmin;Park, Jong-Il;Kim, Yookyung;Kim, Kwang-Yong
    • Proceedings of the Korean Society of Broadcast Engineers Conference
    • /
    • /
    • pp.72-75
    • /
    • 2016
  • 본 논문에서는 고정된 카메라 환경에서 카메라의 흔들림에 강인한 배경 영상을 생성할 수 있는 배경 모델링 방법을 제안한다. 흔들리지 않은 영상을 기준 영상으로 설정하고 기준 영상에서 해리스 코너 검출기를 이용하여 특징점들을 검출한다. 이후 입력 영상에 대해 동일한 방식으로 특징점을 추출한 뒤 탬플릿 매칭과 거리 비교를 이용하여 공통적으로 나타나는 배경 영역들에 대한 특징점만을 선별한다. 기준 영상에서의 특징점과 목표 영상에서의 대응되는 특징점 쌍을 이용하여 보정을 위한 호모그래피 행렬을 계산한다. 이렇게 계산된 보정 행렬을 이용하여 흔들린 목표 영상을 보정하게 된다. 흔들린 영상들을 보정한 후 보정된 영상들로 배경 모델을 생성하게 되면 정확한 배경 모델을 생성할 수 있다.

  • PDF

Transfer Learning using Multiple ConvNet Layers Activation Features with Principal Component Analysis for Image Classification (전이학습 기반 다중 컨볼류션 신경망 레이어의 활성화 특징과 주성분 분석을 이용한 이미지 분류 방법)

  • Byambajav, Batkhuu;Alikhanov, Jumabek;Fang, Yang;Ko, Seunghyun;Jo, Geun Sik
    • Journal of Intelligence and Information Systems
    • /
    • v.24 no.1
    • /
    • pp.205-225
    • /
    • 2018
  • Convolutional Neural Network (ConvNet) is one class of the powerful Deep Neural Network that can analyze and learn hierarchies of visual features. Originally, first neural network (Neocognitron) was introduced in the 80s. At that time, the neural network was not broadly used in both industry and academic field by cause of large-scale dataset shortage and low computational power. However, after a few decades later in 2012, Krizhevsky made a breakthrough on ILSVRC-12 visual recognition competition using Convolutional Neural Network. That breakthrough revived people interest in the neural network. The success of Convolutional Neural Network is achieved with two main factors. First of them is the emergence of advanced hardware (GPUs) for sufficient parallel computation. Second is the availability of large-scale datasets such as ImageNet (ILSVRC) dataset for training. Unfortunately, many new domains are bottlenecked by these factors. For most domains, it is difficult and requires lots of effort to gather large-scale dataset to train a ConvNet. Moreover, even if we have a large-scale dataset, training ConvNet from scratch is required expensive resource and time-consuming. These two obstacles can be solved by using transfer learning. Transfer learning is a method for transferring the knowledge from a source domain to new domain. There are two major Transfer learning cases. First one is ConvNet as fixed feature extractor, and the second one is Fine-tune the ConvNet on a new dataset. In the first case, using pre-trained ConvNet (such as on ImageNet) to compute feed-forward activations of the image into the ConvNet and extract activation features from specific layers. In the second case, replacing and retraining the ConvNet classifier on the new dataset, then fine-tune the weights of the pre-trained network with the backpropagation. In this paper, we focus on using multiple ConvNet layers as a fixed feature extractor only. However, applying features with high dimensional complexity that is directly extracted from multiple ConvNet layers is still a challenging problem. We observe that features extracted from multiple ConvNet layers address the different characteristics of the image which means better representation could be obtained by finding the optimal combination of multiple ConvNet layers. Based on that observation, we propose to employ multiple ConvNet layer representations for transfer learning instead of a single ConvNet layer representation. Overall, our primary pipeline has three steps. Firstly, images from target task are given as input to ConvNet, then that image will be feed-forwarded into pre-trained AlexNet, and the activation features from three fully connected convolutional layers are extracted. Secondly, activation features of three ConvNet layers are concatenated to obtain multiple ConvNet layers representation because it will gain more information about an image. When three fully connected layer features concatenated, the occurring image representation would have 9192 (4096+4096+1000) dimension features. However, features extracted from multiple ConvNet layers are redundant and noisy since they are extracted from the same ConvNet. Thus, a third step, we will use Principal Component Analysis (PCA) to select salient features before the training phase. When salient features are obtained, the classifier can classify image more accurately, and the performance of transfer learning can be improved. To evaluate proposed method, experiments are conducted in three standard datasets (Caltech-256, VOC07, and SUN397) to compare multiple ConvNet layer representations against single ConvNet layer representation by using PCA for feature selection and dimension reduction. Our experiments demonstrated the importance of feature selection for multiple ConvNet layer representation. Moreover, our proposed approach achieved 75.6% accuracy compared to 73.9% accuracy achieved by FC7 layer on the Caltech-256 dataset, 73.1% accuracy compared to 69.2% accuracy achieved by FC8 layer on the VOC07 dataset, 52.2% accuracy compared to 48.7% accuracy achieved by FC7 layer on the SUN397 dataset. We also showed that our proposed approach achieved superior performance, 2.8%, 2.1% and 3.1% accuracy improvement on Caltech-256, VOC07, and SUN397 dataset respectively compare to existing work.

Hierarchical Feature Based Block Motion Estimation for Ultrasound Image Sequences (초음파 영상을 위한 계층적 특징점 기반 블록 움직임 추출)

  • Kim, Baek-Sop;Shin, Seong-Chul
    • Journal of KIISE:Software and Applications
    • /
    • v.33 no.4
    • /
    • pp.402-410
    • /
    • 2006
  • This paper presents a method for feature based block motion estimation that uses multi -resolution image sequences to obtain the panoramic images in the continuous ultrasound image sequences. In the conventional block motion estimation method, the centers of motion estimation blocks are set at the predetermined and equally spaced locations. This requires the large blocks to include at least one feature, which inevitably requires long estimation time. In this paper, we propose an adaptive method which locates the center of the motion estimation blocks at the feature points. This make it possible to reduce the block size while keeping the motion estimation accuracy The Harris-Stephen corner detector is used to get the feature points. The comer points tend to group together, which cause the error in the global motion estimation. In order to distribute the feature points as evenly as Possible, the image is firstly divided into regular subregions, and a strongest corner point is selected as a feature in each subregion. The ultrasound Images contain speckle patterns and noise. In order to reduce the noise artifact and reduce the computational time, the proposed method use the multi-resolution image sequences. The first algorithm estimates the motion in the smoothed low resolution image, and the estimated motion is prolongated to the next higher resolution image. By this way the size of search region can be reduced in the higher resolution image. Experiments were performed on three types of ultrasound image sequences. These were shown that the proposed method reduces both the computational time (from 77ms to 44ms) and the displaced frame difference (from 66.02 to 58.08).

FPGA Design of SVM Classifier for Real Time Image Processing (실시간 영상처리를 위한 SVM 분류기의 FPGA 구현)

  • Na, Won-Seob;Han, Sung-Woo;Jeong, Yong-Jin
    • Journal of IKEEE
    • /
    • v.20 no.3
    • /
    • pp.209-219
    • /
    • 2016
  • SVM is a machine learning method used for image processing. It is well known for its high classification performance. We have to perform multiple MAC operations in order to use SVM for image classification. However, if the resolution of the target image or the number of classification cases increases, the execution time of SVM also increases, which makes it difficult to be performed in real-time applications. In this paper, we propose an hardware architecture which enables real-time applications using SVM classification. We used parallel architecture to simultaneously calculate MAC operations, and also designed the system for several feature extractors for compatibility. RBF kernel was used for hardware implemenation, and the exponent calculation formular included in the kernel was modified to enable fixed point modelling. Experimental results for the system, when implemented in Xilinx ZC-706 evaluation board, show that it can process 60.46 fps for $1360{\times}800$ resolution at 100MHz clock frequency.