In this paper, we present Hybrid MAC (H-MAC), a novel low power with minimal packet delay medium access control protocol for wireless sensor networks (WSNs). H-MAC achieves high energy efficiency under wide range of traffic load. It ensures high channel utilization during high traffic load without compromising energy efficiency. H-MAC does it by using the strength of CSMA and TDMA approach with intelligence. The novel idea behind the H-MAC is that, it uses both the broadcast scheduling and link scheduling. Depending on the network loads the H-MAC protocol dynamically switches from broadcast scheduling to link scheduling and vice-versa in order to achieve better efficiency. Furthermore, H-MAC uses Request-To-Send (RTS), Clear‒To-send (CTS) handshakes with methods for adapting the transmit power to the minimum level necessary to reach the intended neighbor with a given BER target or packet loss probability. Thus H-MAC reduces energy consumption by suitably varying the transmit power. The simulation results corroborate the theoretical idea, and show the efficiency of our proposed protocol.

抄録全体を表示

There have been no confirmed efficient fishery investigation methods for Laminaria beds in Nemuro, Hokkaido, Japan. Recently, using an underwater video camera has become an increasingly common way to investigate Laminaria beds. However, data obtained from the investigation need to be manually analyzed, which is inefficient. To develop an automated system for measuring the amount of fishery resources from images taken with an underwater camera for fishery investigation of Laminaria beds in Nemuro, Hokkaido, Japan. Ezo-wakame (a type of seaweed) extraction method using image analysis technology has been developed and evaluated in an experiment. In this paper, we explain about extraction method of Ezo-wakame and the experiment of applying the proposed method. This method uses the vein on the surface of Ezo-wakame to extract it because veins are specific characteristics and indicate the number of Ezo-wakames.

抄録全体を表示

Spatial pyramid matching (SPM) has been an important approach to image categorization. This method partitions the image into increasingly fine sub-regions and computes histograms of local features at each sub-region. Although SPM is an efficient extension of an unordered bag-of-features image representation, it still measures the similarity between sub-regions by application of the bag-of-features model. Therefore, it is limited in its capacity to achieve optimal matching between sets of unordered features. To overcome this limitation, we propose a hierarchical spatial matching kernel (HSMK) that uses a coarse-to-fine model for the sub-regions to obtain better optimal matching approximations. Our proposed kernel can deal robustly with unordered feature sets as well as various cardinalities. In experiments, results of HSMK outperformed those of SPM and led to state-of-the-art performance on several well-known databases of benchmarks in image categorization, even though we use only a single type of image feature.

抄録全体を表示

本論文では検出履歴学習による自動最適化テンプレートを用いた高速画像マッチングを提案する．テンプレートマッチングは物体検出の一手法として数多くの手法が提案されているが，その中に機械学習によってテンプレートを最適化する手法がある．しかし，それらの手法は大量のデータを学習させる必要があるため，テンプレート画像を最適化するために長い前処理の時間を必要とする．さらに，機械学習のためのデータの準備も容易ではない．この問題を解決するために，本稿では検出と同時に検出履歴を学習してテンプレート画像を最適化する手法を提案する．実験により，大量のデータが前もって準備されていなくてもテンプレート画像を最適化できることを示す．

抄録全体を表示

動物行動学では，群れの行動を解析するために密度推定を行う必要がある．群れは数個体から数千個体の規模で形成されるものなど様々である．このような群れの行動を記録した映像の解析には画像処理による自動化が必要である．一方で画像処理において，色彩や形状的特徴に乏しい複数の類似物体を個体追跡することは困難である．本論文では，ミナミコメツキガニの群行動解析を目的とし，視覚的特徴が得られない複数類似物体を個体追跡するための密な隣接，重なりに対応できる手法を提案している．具体的には背景差分法により移動物体領域を抽出した後，各移動物体領域から脚領域を収縮・膨張処理により除去し，個体候補領域に対してな隣接や重なりを考慮した評価関数を用いて分布推定を行う．提案手法をシミュレーション実験と実際のミナミコメツキガニの映像を用いた群行動解析実験に適用した結果，従来手法に対しその有効性が検証できた．

抄録全体を表示

本論文では，Fast Level Set Method (FLSM)に基づいた三次元形状復元において，2体の人体間における相互オクルージョン発生にも対応可能な人体形状推定手法を提案する．従来のFLSMを用いた手法では，2体の人体間に相互オクルージョン生じた際，FLSMに対し拘束を与えるため，相互オクルージョン発生状態で人体の姿勢が変化した場合，正確な人体形状復元を行えない．そこで，提案手法では，相互オクルージョン発生時における人体の概形領域を直接求め，FLSMによる特別な拘束を必要としない人休概形復元を実現する．さらに，復元した人体概形から抽出した姿勢情報をもとに，相互オクルージョン発生寸前にFLSMで復元した人体形状のパーツを再構成することで，より実際の人体に近い形状推定を実現する．

抄録全体を表示

This paper proposes a low power video decoding with adaptive granularity in temporal scalability. This proposal can be applied to reduce the computational complexity of H.264/AVC decoder with acceptable loss of the video quality, and make the single layer bit stream sources much more flexible for various terminal devices. Proposed low power decoding process consists of four proposed algorithms, the reference frame index decision algorithm, motion vector composition algorithm, block-partition decision algorithm and the adaptive selecting algorithm for skipped frames. The experiment results show that the reduction rate of the decoding time decreases when the number of skipped frames increases, and the loss of the video quality increases at the same time. The PSNR loss in the B frame skipping is much smaller than the PSNR loss in the P frame skipping. In the fixed frame skipping cases, the 2/3 P frame is skipped with 60% decoding time reduction and 2.73 dB average PSNR loss in the all filling comparison or 1.59 dB average PSNR loss in the corresponding comparison. Analyzing the relation between motion vector information and the video quality loss of the corresponding frames in probability shows that the proposed adaptive skipping scheme reduces quality loss by skipping the frames with slight movements and keeping the frames with strong movements. Based on the adaptive skipping scheme, the average PSNR is improved 0.68 dB in the all filling comparison or 0.60 dB in the corresponding comparison compared with the fixed frame skipping scheme with almost the same reduction rate of the decoding time.

抄録全体を表示