Perspective-SIFT: An efficient tool for low-altitude remote sensing image registration

Cai, Guorong; Jodoin, Pierre-Marc; Li, Shaozi; Wu, Yundong; Su, Songzhi; Huang, Zhenkun

doi:10.1016/j.sigpro.2013.04.008

Cited by 58 publications

(32 citation statements)

References 43 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the match stage, feature extraction and matching, such as PSIFT [55] along with kd-tree, are conducted on two UAV aerial images. Parallax of the two images should larger than a certain threshold to ensure that keypoint coordinates in two images are not identical.…”

Section: The Proposed Residual Analysis Methodsmentioning

confidence: 99%

A Robust Transform Estimator Based on Residual Analysis and Its Application on UAV Aerial Images

Cai

Leng

et al. 2018

Remote Sensing

View full text Add to dashboard Cite

Abstract:Estimating the transformation between two images from the same scene is a fundamental step for image registration, image stitching and 3D reconstruction. State-of-the-art methods are mainly based on sorted residual for generating hypotheses. This scheme has acquired encouraging results in many remote sensing applications. Unfortunately, mainstream residual based methods may fail in estimating the transform between Unmanned Aerial Vehicle (UAV) low altitude remote sensing images, due to the fact that UAV images always have repetitive patterns and severe viewpoint changes, which produce lower inlier rate and higher pseudo outlier rate than other tasks. We performed extensive experiments and found the main reason is that these methods compute feature pair similarity within a fixed window, making them sensitive to the size of residual window. To solve this problem, three schemes that based on the distribution of residuals are proposed, which are called Relational Window (RW), Sliding Window (SW), Reverse Residual Order (RRO), respectively. Specially, RW employs a relaxation residual window size to evaluate the highest similarity within a relaxation model length. SW fixes the number of overlap models while varying the length of window size. RRO takes the permutation of residual values into consideration to measure similarity, not only including the number of overlap structures, but also giving penalty to reverse number within the overlap structures. Experimental results conducted on our own built UAV high resolution remote sensing images show that the proposed three strategies all outperform traditional methods in the presence of severe perspective distortion due to viewpoint change.

show abstract

Section: The Proposed Residual Analysis Methodsmentioning

confidence: 99%

A Robust Transform Estimator Based on Residual Analysis and Its Application on UAV Aerial Images

Cai

Leng

et al. 2018

Remote Sensing

View full text Add to dashboard Cite

show abstract

“…In this study, a piecewise function is adopted to adjust the learning rate. The initial learning rate is set to 0.01 then decreased gradually with the following formula: (17) in which iter denotes the number of iterations; η iter denotes the learning rate of the iter th iteration, which is updated based on previous learning rate η iter−1 ; % is an operator for computing the remainder; the optimal convergence can be achieved by decreasing the learning rate at about every 100 iterations based on the observation of our experiments; and α is a constant, which is set to 0.75. Figure 8 shows the visualization of features at each convolutional layer (Conv1-Conv6) of the SCNN after ReLU activation.…”

Section: Scnn Trainingmentioning

confidence: 99%

“…Many improved descriptors, such as principal component analysis-SIFT [11], gradient location and orientation histogram [12], and Affine-SIFT [13], have been investigated to make the SIFT features distinctive in image deformation. Feature descriptors are combined with several similarity metrics or constraints, such as scale-orientation joint restriction criteria [14], weight-based topological map-matching algorithm [15], normalized cross correlation and least square matching [16], perspective scale invariant feature [17], l q -estimator [18], and L 2 -minimizing estimation [6], to match remote sensing images. Despite significant improvements to the feature-based matching method, the manually designed methods (e.g., SIFT) cannot fully obtain the invariant descriptors with the appearance of nonlinear illumination changes, shadows, and occlusions [19].…”

Section: Introductionmentioning

confidence: 99%

Matching of Remote Sensing Images with Complex Background Variations via Siamese Convolutional Neural Network

Chen

et al. 2018

Remote Sensing

View full text Add to dashboard Cite

Abstract:Feature-based matching methods have been widely used in remote sensing image matching given their capability to achieve excellent performance despite image geometric and radiometric distortions. However, most of the feature-based methods are unreliable for complex background variations, because the gradient or other image grayscale information used to construct the feature descriptor is sensitive to image background variations. Recently, deep learning-based methods have been proven suitable for high-level feature representation and comparison in image matching. Inspired by the progresses made in deep learning, a new technical framework for remote sensing image matching based on the Siamese convolutional neural network is presented in this paper. First, a Siamese-type network architecture is designed to simultaneously learn the features and the corresponding similarity metric from labeled training examples of matching and non-matching true-color patch pairs. In the proposed network, two streams of convolutional and pooling layers sharing identical weights are arranged without the manually designed features. The number of convolutional layers is determined based on the factors that affect image matching. The sigmoid function is employed to compute the matching and non-matching probabilities in the output layer. Second, a gridding sub-pixel Harris algorithm is used to obtain the accurate localization of candidate matches. Third, a Gaussian pyramid coupling quadtree is adopted to gradually narrow down the searching space of the candidate matches, and multiscale patches are compared synchronously. Subsequently, a similarity measure based on the output of the sigmoid is adopted to find the initial matches. Finally, the random sample consensus algorithm and the whole-to-local quadratic polynomial constraints are used to remove false matches. In the experiments, different types of satellite datasets, such as ZY3, GF1, IKONOS, and Google Earth images, with complex background variations are used to evaluate the performance of the proposed method. The experimental results demonstrate that the proposed method, which can significantly improve the matching performance of multi-temporal remote sensing images with complex background variations, is better than the state-of-the-art matching methods. In our experiments, the proposed method obtained a large number of evenly distributed matches (at least 10 times more than other methods) and achieved a high accuracy (less than 1 pixel in terms of root mean square error).

show abstract

“…SIFT locates key point by finding scale-space extreme in the difference-of-Gaussian (DoG) function; once a key point is detected, it will be described and matched based on its orientation, scale and coordination. SIFT and its variations are the most popular and successful feature detection algorithms in many remote sensing registration applications [2,9,[13][14][15][16]. However, if the image pairs are obtained under a large difference of camera viewpoints, the correspondences generated by SIFT would be too few to perform a reliable registration.…”

Section: Introductionmentioning

confidence: 99%

An ASIFT-Based Local Registration Method for Satellite Imagery

et al. 2015

View full text Add to dashboard Cite

Imagery registration is a fundamental step, which greatly affects later processes in image mosaic, multi-spectral image fusion, digital surface modelling, etc., where the final solution needs blending of pixel information from more than one images. It is highly desired to find a way to identify registration regions among input stereo image pairs with high accuracy, particularly in remote sensing applications in which ground control points (GCPs) are not always available, such as in selecting a landing zone on an outer space planet. In this paper, a framework for localization in image registration is developed. It strengthened the local registration accuracy from two aspects: less reprojection error and better feature point distribution. Affine scale-invariant feature transform (ASIFT) was used for acquiring feature points and correspondences on the input images. Then, a homography matrix was estimated as the transformation model by an improved random sample consensus (IM-RANSAC) algorithm. In order to identify a registration region with a better spatial distribution of feature points, the Euclidean distance between the feature points is applied (named the S criterion). Finally, the parameters of the homography matrix were optimized by the Levenberg-Marquardt (LM) algorithm with selective feature points from the chosen registration region. In the experiment section, the Chang'E-2 satellite remote sensing imagery was used for evaluating the performance of the proposed method. The experiment result demonstrates that the proposed method can automatically locate a specific region with high registration accuracy between input images by achieving lower root mean square error (RMSE) and better distribution of feature points.

show abstract

Perspective-SIFT: An efficient tool for low-altitude remote sensing image registration

Cited by 58 publications

References 43 publications

A Robust Transform Estimator Based on Residual Analysis and Its Application on UAV Aerial Images

A Robust Transform Estimator Based on Residual Analysis and Its Application on UAV Aerial Images

Matching of Remote Sensing Images with Complex Background Variations via Siamese Convolutional Neural Network

An ASIFT-Based Local Registration Method for Satellite Imagery

Contact Info

Product

Resources

About