This paper addresses the problem of fitting a functional model to data corrupted with outliers using a multilayered feed-forward neural network (MFNN). Almost all previous efforts to solve this problem have focused on...
详细信息
ISBN:
(纸本)9781424496365
This paper addresses the problem of fitting a functional model to data corrupted with outliers using a multilayered feed-forward neural network (MFNN). Almost all previous efforts to solve this problem have focused on using a training algorithm that minimizes an M-estimator based error criterion. However the robustness gained from M-estimators is still low. Using a training algorithm based on the RANdom SAmple Consensus (ransac) framework improves significantly the robustness of the algorithm. However the algorithm typically requires prolonged period of time before a final solution is reached. In this paper, we propose a new strategy to improve the time performance of the ransac algorithm for training MFNNs. A statistical pre-test based on Wald's sequential probability ratio test (SPRT) is performed on each randomly generated sample to decide whether it deserves to be used for model estimation. The proposed algorithm is evaluated on synthetic data, contaminated with varying degrees of outliers, and have demonstrated faster performance compared to the original ransac algorithm with no significant sacrifice of the robustness.
Subgrade sea and sky monitoring equipment requires the accurate detection of threat targets in a given area. Due to the extremely complex sea-land-sky backgrounds, the sea-sky line is often submerged in the background...
详细信息
Subgrade sea and sky monitoring equipment requires the accurate detection of threat targets in a given area. Due to the extremely complex sea-land-sky backgrounds, the sea-sky line is often submerged in the background. Therefore, we propose an algorithm for accurately detecting sea-sky lines under a complex sea-land-sky background. Based on the analysis of infrared images with sea-land-sky backgrounds, we segment images using the k-means algorithm. To use the random sampling consistency (ransac) algorithm to fit the sea-sky line better, we divide the images into nonuniform segments and count the row mean-value gradient trough. The experimental results show that the sea-sky line can be detected in 1215 out of 1227 pictures. The test success rate is 99%, and the difference from the actual sea-sky line is less than 3 pixels. The method presented has higher adaptability under a subgrade sea and sky monitoring environment.
In Ostia, the huge range of excavation carried out by Guido Calza under Mussolini (1938-1942), the zone of contiguous city blocks unearthed in those massive campaigns. From 2012, new survey by a Japanese team of stand...
详细信息
In Ostia, the huge range of excavation carried out by Guido Calza under Mussolini (1938-1942), the zone of contiguous city blocks unearthed in those massive campaigns. From 2012, new survey by a Japanese team of standing remains using laser scanners formed the basis for an analysis of building history, and for a reconstruction of the original building. There is a considerable amount of undocumented reconstruction work in the upper part of the structure which has been identified from analysis of the surface of the walls. The seam and the absence of coursing between the original walls and the later restored works sometimes including in the Roman phase, and sometimes modern using original part of the walls, make difficult to identify which part of walls were original and which were restorations or re-use 80 years later from the excavation. In this paper, the case that the seams are invisible, but its existence is known from the photographic record of the progress of the excavations. The detection by applying the ransac algorithm to the point cloud data from laser scanning relies on several cases of invisible seams running on the surfaces. Additionally, this method allows us without any special knowledge and experience to find detailed characteristics on the surface of the walls, such as slight unevenness or weathering parts, to extrapolate the building history.
The traditional lane detection methods based on the ransac algorithm used to cause many false detection and unable to accurately detect the lanes in complex road environment, because of the existence of interferential...
详细信息
ISBN:
(纸本)9781538613863
The traditional lane detection methods based on the ransac algorithm used to cause many false detection and unable to accurately detect the lanes in complex road environment, because of the existence of interferential noise points in the set of sampling points. Aiming at these issues, this paper presents a new lane detection method combined fuzzy control with ransac algorithm. The first process of the new lane detection method is pretreatment, the purpose of which is to denoise the image preliminarily through filtering and binarization. And then it selects the region of interest (ROI) that contains lanes in the input image and extract the initial boundary candidate points of the lanes in ROI. So far, there are still a lot of irrelevant noise points in the set of lane boundary candidate points. It would analyze the relationship between the interferential noise points and the boundary points of the lane, and then remove the interferential noise points from the set of lane boundary candidate point by using the fuzzy control. After that, fit the lane model by using ransac algorithm in the set of effective lane boundary points. The experiment shows that the method proposed in this paper has high robustness and effectiveness which can accurately detect the lanes in complicated city road.
An improved ransac algorithm based on structural similarity was proposed to improve the speed and accuracy of traditional ransac (Random Sample Consensus) algorithm and to reduce iterations and runtime. Firstly, BRISK...
详细信息
ISBN:
(纸本)9781509041558
An improved ransac algorithm based on structural similarity was proposed to improve the speed and accuracy of traditional ransac (Random Sample Consensus) algorithm and to reduce iterations and runtime. Firstly, BRISK (Binary Robust Invariant Scalable Keypoints) algorithm was used to extract and describe feature points. The initial match set is obtained by hamming distance feature matching. Then, mismatches are eliminated by similar structure constraints. Finally, the new match set is taken as the input of ransac to calculate the transformation matrix. The algorithm can obtain the transformation model quickly because it has purified matching points after the initial matching. Experiments show that the number of iterations and runtime of this algorithm are obviously less than the number of iterations and runtime of the traditional algorithm. So the proposed method outperforms traditional ransac (Random Sample Consensus) algorithm significantly both in iterations and speed.
Aiming at the problems that the current ransac (Random Sample Consensus) algorithm has too large randomness and is susceptible to external point interference, which leads to the reduction of matching accuracy, an impr...
详细信息
ISBN:
(纸本)9783031138355;9783031138348
Aiming at the problems that the current ransac (Random Sample Consensus) algorithm has too large randomness and is susceptible to external point interference, which leads to the reduction of matching accuracy, an improved ransac algorithm combining feature matching confidence and grid clustering is proposed. Firstly, rough matching is carried out by the FLANN algorithm, and confidence analysis is carried out on the coarse matching point pairs, then expanding grid clustering around the high confidence point pairs. Multiple local optimal interior points are screened to optimize the global interior points and improve the matching accuracy of feature points. The experimental results show that the improved ransac in this paper increases the existence probability of interior points, avoids too many wrong feature matching affecting the model effect of the homography matrix, and improves the accuracy of feature matching.
The computer vision involves many modeling problems with preventing noise caused by disturbance and sensing unit conditions. In order to improve computer vision system performance, a robust modeling technique must be ...
详细信息
ISBN:
(纸本)9781728165417
The computer vision involves many modeling problems with preventing noise caused by disturbance and sensing unit conditions. In order to improve computer vision system performance, a robust modeling technique must be developed for essential models in the system. The random sample consensus (ransac) and least median of squares (LMedS) algorithm have been widely applied in such issues. However, the performance deteriorates as the noise ratio increases and the modeling time for algorithms tends to increase in industrial applications. As an effective technique, we proposed a new fuzzy ransac method based on reinforcement learning concept for robust modeling. In this study, we proposed a new technique for the fuzzy ransac in order to improve learning performance in initial learning stage based on weighted modeling technique. Through modeling synthetic nonlinear data and camera homography experiments, the performance of the technique was evaluated. Their results found the proposed technique to be promising for improving modeling performance in initial learning stage.
This study attempts a solution for autonomous vehicles to avoid immediate collision due to close proximity between cars. Since LIDAR sensors are widely used for capturing images in autonomous car industry, we depict a...
详细信息
ISBN:
(纸本)9783319990071;9783319990064
This study attempts a solution for autonomous vehicles to avoid immediate collision due to close proximity between cars. Since LIDAR sensors are widely used for capturing images in autonomous car industry, we depict a scope of using ransac algorithm and linear regression to reconstruct the orthoimages to escape traffic bottleneck as well as avoid collision. It is found that LIDAR sensors can't suggests much detail in close distance, and cameras don't perform well in conditions with low light or glare images. Dataset is collected from KITTI (Karlsruhe Institute of Technology) containing compressed pixels. Significance resultants focus on error reduction followed by feature extraction simulated with MATLAB. The findings excludes large scale of data size to implement and project in T-way testing for determining strength as well as capability of resultants.
The computer vision involves many modeling problems with preventing noise caused by disturbance and sensing unit conditions. In order to improve computer vision system performance, a robust modeling technique must be ...
详细信息
The computer vision involves many modeling problems with preventing noise caused by disturbance and sensing unit conditions. In order to improve computer vision system performance, a robust modeling technique must be developed for essential models in the system. The random sample consensus (ransac) and least median of squares (LMedS) algorithm have been widely applied in such issues. However, the performance deteriorates as the noise ratio increases and the modeling time for algorithms tends to increase in industrial applications. As an effective technique, we proposed a new fuzzy ransac method based on reinforcement learning concept for robust modeling. In this study, we proposed a new technique for the fuzzy ransac in order to improve learning performance in initial learning stage based on weighted estimation technique. Through modeling synthetic nonlinear data and camera homography experiments, the performance of the technique was evaluated. Their results found the proposed technique to be promising for improving modeling performance in initial learning stage.
In this study, a method to enhance the accuracy of overlapped etched track analysis is proposed. Counting tracks by eye is not an easy task and automated tracks counting systems are attractive key for this problem. Th...
详细信息
In this study, a method to enhance the accuracy of overlapped etched track analysis is proposed. Counting tracks by eye is not an easy task and automated tracks counting systems are attractive key for this problem. This method supplements the deficiencies of the conventional track analysis method. A computer programme named KoreaTech Track Measurement System written in C++, which is the authors' previous method, has been upgraded. In the proposed track analysis method, the track images captured from solid state nuclear track detectors are geometrically analysed and the number of tracks is counted. A damaged etching track shape can be restored on the track image to improve the analysis accuracy. For track restoration, the effective points are differentiated from the damaged track image. The track image is then restored by estimating the radii (small object removal) or their axis (ellipse, circle and non-circle) using the RANdom sample consensus method. Using the restored track image, the track parameters are obtained from the ellipse and then approximated to the contour of the track image to analyse the track image. Then, the total number of tracks including the overlapped tracks is counted. To verify the proposed track analysis method, experiments using actual etching track images are conducted and the results are discussed.
暂无评论