Aerial image sequence mosaicking is one of the chal-lenging research fields in computer vision.To obtain large-scale orthophoto maps with object detection information,we propose a vision-based image mosaicking algorit...Aerial image sequence mosaicking is one of the chal-lenging research fields in computer vision.To obtain large-scale orthophoto maps with object detection information,we propose a vision-based image mosaicking algorithm without any extra location data.According to object detection results,we define a complexity factor to describe the importance of each input ima-ge and dynamically optimize the feature extraction process.The feature points extraction and matching processes are mainly guided by the speeded-up robust features(SURF)and the grid motion statistic(GMS)algorithm respectively.A robust refer-ence frame selection method is proposed to eliminate the trans-formation distortion by searching for the center area based on overlaps.Besides,the sparse Levenberg-Marquardt(LM)al-gorithm and the heavy occluded frames removal method are ap-plied to reduce accumulated errors and further improve the mo-saicking performance.The proposed algorithm is performed by using multithreading and graphics processing unit(GPU)accel-eration on several aerial image datasets.Extensive experiment results demonstrate that our algorithm outperforms most of the existing aerial image mosaicking methods in visual quality while guaranteeing a high calculation speed.展开更多
Define the incremental fractional Brownian field with parameter H ∈ （0, 1） by ZH（τ, s） = BH（s-＋τ） - BH（S）, where BH（s） is a fractional Brownian motion with Hurst parameter H ∈ （0, 1）. We firstly deriv...Define the incremental fractional Brownian field with parameter H ∈ （0, 1） by ZH（τ, s） = BH（s-＋τ） - BH（S）, where BH（s） is a fractional Brownian motion with Hurst parameter H ∈ （0, 1）. We firstly derive the exact tail asymptoties for the maximum MH＊（T） = max（τ,s）∈[a,b]×[0,T] ZH（τ, s）/τH of the standardised fractional Brownian motion field, with any fixed 0 〈 a 〈 b 〈 ∞ and T 〉 0; and we, furthermore, extend the obtained result to the ease that T is a positive random variable independent of {BH（s）, s ≥ 0}. As a by-product, we obtain the Gumbel limit law for MH＊r（T） as T →∞.展开更多
The problem of recognizing natural scenes, such as water, smoke, fire, wind-blown vegetation and a flock of flying birds, is considered. These scenes exhibit the characteristic dynamic pattern, but have stochastic ext...The problem of recognizing natural scenes, such as water, smoke, fire, wind-blown vegetation and a flock of flying birds, is considered. These scenes exhibit the characteristic dynamic pattern, but have stochastic extent. They are referred to as dynamic texture（DT）. In reality, the diversity of DTs on different viewpoints and scales are very common, which also bring great difficulty to recognize DTs. In the previous studies, due to no considering of the deformable and transient nature of elements in DT, the motion estimation method is based on brightness constancy assumption,which seem inappropriate for aggregate and complex motions. A novel motion model based on relative motion in the neighborhood of two-dimensional motion fields is proposed. The estimation of non-rigid motion of DTs is based on the continuity equation, and then the local vector difference（LVD） is proposed to characterize DT local relative motion. Spatiotemporal statistics of the LVDs is used as the representation of DT sequences. Excellent performances of classifying all DTs in UCLA database demonstrate the capability of the proposed method in describing DT.展开更多
基金supported by the National Natural Science Foundation of China(6160304061973036).
文摘Aerial image sequence mosaicking is one of the chal-lenging research fields in computer vision.To obtain large-scale orthophoto maps with object detection information,we propose a vision-based image mosaicking algorithm without any extra location data.According to object detection results,we define a complexity factor to describe the importance of each input ima-ge and dynamically optimize the feature extraction process.The feature points extraction and matching processes are mainly guided by the speeded-up robust features(SURF)and the grid motion statistic(GMS)algorithm respectively.A robust refer-ence frame selection method is proposed to eliminate the trans-formation distortion by searching for the center area based on overlaps.Besides,the sparse Levenberg-Marquardt(LM)al-gorithm and the heavy occluded frames removal method are ap-plied to reduce accumulated errors and further improve the mo-saicking performance.The proposed algorithm is performed by using multithreading and graphics processing unit(GPU)accel-eration on several aerial image datasets.Extensive experiment results demonstrate that our algorithm outperforms most of the existing aerial image mosaicking methods in visual quality while guaranteeing a high calculation speed.
基金supported by National Natural Science Foundation of China(Grant Nos.11326175 and 71471090)Natural Science Foundation of Zhejiang Province of China(Grant No.LQ14A010012)+2 种基金Research Start-up Foundation of Jiaxing University(Grant No.70512021)China Postdoctoral Science Foundation(Grant No.2014T70449)Natural Science Foundation of Jiangsu Province of China(Grant No.BK20131339)
文摘Define the incremental fractional Brownian field with parameter H ∈ （0, 1） by ZH（τ, s） = BH（s-＋τ） - BH（S）, where BH（s） is a fractional Brownian motion with Hurst parameter H ∈ （0, 1）. We firstly derive the exact tail asymptoties for the maximum MH＊（T） = max（τ,s）∈[a,b]×[0,T] ZH（τ, s）/τH of the standardised fractional Brownian motion field, with any fixed 0 〈 a 〈 b 〈 ∞ and T 〉 0; and we, furthermore, extend the obtained result to the ease that T is a positive random variable independent of {BH（s）, s ≥ 0}. As a by-product, we obtain the Gumbel limit law for MH＊r（T） as T →∞.
基金supported by the National Natural Science Foundation of China(41504115)the Shaanxi Province Natural Science Foundation(2015JQ6223)+2 种基金the Foundation of Strengthening Police Science and Technology from Ministry of Public Security(2015GABJC50)the International Technology Cooperation Plan Project of Shaanxi Province(2015KW-0142015KW-013)
文摘The problem of recognizing natural scenes, such as water, smoke, fire, wind-blown vegetation and a flock of flying birds, is considered. These scenes exhibit the characteristic dynamic pattern, but have stochastic extent. They are referred to as dynamic texture（DT）. In reality, the diversity of DTs on different viewpoints and scales are very common, which also bring great difficulty to recognize DTs. In the previous studies, due to no considering of the deformable and transient nature of elements in DT, the motion estimation method is based on brightness constancy assumption,which seem inappropriate for aggregate and complex motions. A novel motion model based on relative motion in the neighborhood of two-dimensional motion fields is proposed. The estimation of non-rigid motion of DTs is based on the continuity equation, and then the local vector difference（LVD） is proposed to characterize DT local relative motion. Spatiotemporal statistics of the LVDs is used as the representation of DT sequences. Excellent performances of classifying all DTs in UCLA database demonstrate the capability of the proposed method in describing DT.
