AI小作坊 的博客

The sources of error for the edge finding technique proposed by Marr and Hildreth (D. Marr and T. Poggio, Proc. R. Soc. London Ser. B204, 1979, 301–328; D. Marr and E. Hildreth, Proc. R. Soc. London Ser. B.207, 1980, 187–217) are identified, and the magnitudes of the errors are estimated, based on idealized models of the most common error producing situations. Errors are shown to be small for linear illuminations, as well as for nonlinear illuminations with a second derivative less than a critical value. Nonlinear illuminations are shown to lead to spurious contours under some conditions, and some fast techniques for discarding such contours are suggested.

Canny (1986) suggested that an optimal edge detector should maximize both signal-to-noise ratio and localization, and he derived mathematical expressions for these criteria. Based on these criteria, he claimed that the optimal step edge detector was similar to a derivative of a gaussian. However, Canny’s work suffers from two problems. First, his derivation of localization criterion is incorrect. Here we provide a more acurate localization criterion and derive the optimal detector from it. Second, and more seriously, the Canny criteria yield an infinitely wide optimal edge detector. The width of the optimal detector can however be limited by considering the effect of the neighbouring edges in the image. If we do so, we find that the optimal step edge detector, according to the Canny criteria, is the derivative of an ISEF filter, proposed by Shen and Castan (1992). In addition, if we also consider detecting blurred (or non-sharp) gaussian edges of different widths, we find that the optimal blurred-edge detector is the above optimal step edge detector convolved with a gaussian. This implies that edge detection must be performed at multiple scales to cover all the blur widths in the image. We derive a simple scale selection procedure for edge detection, and demonstrate it in one and two dimensions.

Overview of OpenCV 2 Computer Vision Application Programming Cookbook Teaches you how to program computer vision applications in C++ using the different features of the OpenCV library Demonstrates the important structures and functions of OpenCV in detail with complete working examples Describes fundamental concepts in computer vision and image processing Gives you advice and tips to create more effective object-oriented computer vision programs Contains examples with source code and shows results obtained on real images with detailed explanations and the required screenshots

Learning based Symmetric Features Selection for Vehicle Detection This paper describes a symmetric features selection strategy based on statistical learning method for detecting vehicles with a single moving camera for autonomous driving. Symmetry is a good class of feature for vehicle detection, but the areas with high symmetry and threshold for segmentation is hard to be decided. Usually, the additional supposition is added artificially, and this will decrease the robustness of algorithms. In this paper, we focus on the problem of symmetric features selection using learning method for autonomous driving environment. Global symmetry and local symmetry are defined and used to construct a cascaded structure with a one-class classifier followed by a two-class classifier.

Intensity and Edge-Based Symmetry Detection Applied to Car-Following We present two methods for detecting symmetry in images, one based directly on the intensity values and another one based on a discrete representation of local orientation. A symmetry finder has been developed which uses the intensity-based method to search an image for compact regions which display some degree of mirror symmetry due to intensity similarities across a straight axis. In a different approach, we look at symmetry as a bilateral relationship between local orientations. A symmetryenhancing edge detector is presented which indicates edges dependent on the orientations at two different image positions. SEED, as we call it, is a detector element implemented by a feedforward network that holds the symmetry conditions. We use SEED to find the contours of symmetric objects of which we know the axis of symmetry from the intensity-based symmetry finder. The methods presented have been applied to the problem of visually guided car-following. Real-time experiments with a system for automatic headway control on motorways have been successful.

Accurate Robust Symmetry Estimation Stephen Smith and Mark Jenkinson There are various applications, both in medical and non-medical image analysis, which require the automatic detection of the line (2D images) or plane (3D) of reflective symmetry of objects. There exist relatively simple methods of finding reflective symmetry when object images are complete (i.e., completely symmetric and perfectly segmented from image “background”). A much harder problem is finding the line or plane of symmetry when the object of interest contains asymmetries, and may not have well defined edges.

Approach of vehicle segmentation based on texture character

Method of removing moving shadow based on texture

Most video surveillance systems require to manually set a motion detection sensitivity level to generate motion alarms. The performance of motion detection algorithms, embedded in closed circuit television (CCTV) camera and digital video recorder (DVR), usually depends upon the preselected motion sensitivity level, which is expected to work in all environmental conditions. Due to the preselected sensitivity level, false alarms and detection failures usually exist in video surveillance systems. The proposed motion detection model based upon variational energy provides a robust detection method at various illumination changes and noise levels of image sequences without tuning any parameter manually. We analyze the structure mathematically and demonstrate the effectiveness of the proposed model with numerous experiments in various environmental conditions. Due to the compact structure and efficiency of the proposed model, it could be implemented in a small embedded system.

Otsu’s method of image segmentation selects an optimum threshold by maximizing the between-class variance in a gray image. However, this method becomes very time-consuming when extended to a multi-level threshold problem due to the fact that a large number of iterations are required for computing the cumulative probability and the mean of a class. To greatly improve the efficiency of Otsu’s method, a new fast algorithm called the TSMO method (Two-Stage Multithreshold Otsu method) is presented. The TSMO method outperforms Otsu’s method by greatly reducing the iterations required for computing the between-class variance in an image. The experimental results show that the computational time increases exponentially for the conventional Otsu method with an average ratio of about 76. For TSMO-32, the maximum computational time is only 0.463 s when the class number M increases from two to six with relative errors of less than 1% when compared to Otsu’s method. The ratio of computational time of Otsu’s method to TSMO-32 is rather high, up to 109,708, when six classes (M = 6) in an image are used. This result indicates that the proposed method is far more efficient with an accuracy equivalent to Otsu’s method. It also has the advantage of having a small variance in runtimes for different test images.

In this paper, we show a new method to reconstruct and update the background. This approach is based on double-background. We use the statistical information of the pixel intensity to construct a background that represents the status during a long time, and construct another background with feedback information in motion detection that represents the recent changes at a short time. This couple of background images is fused to construct and update the background image used to motion detection. The background reconstruction algorithm can perform well on the tests that we have applied it to.

In this paper we present a statistical change detection approach aimed at being robust with respect to the main disturbance factors acting in real-world applications, such as illumination changes, camera gain and exposure variations, noise. We rely on modeling the effects of disturbance factors on images as locally order-preserving transformations of pixel intensities plus additive noise. This allows us to identify within the space of all the possible image change patterns the subspace corresponding to disturbance factors effects. Hence, scene changes can be detected by a-contrario testing the hypothesis that the measured pattern is due to disturbance factors, that is by computing a distance between the pattern and the subspace. By assuming additive gaussian noise, the distance can be computed within a maximum likelihood non-parametric isotonic regression framework. In particular, the projection of the pattern onto the subspace is computed by an O(N) iterative procedure known as Pool Adjacent Violators algorithm.

Cooperative Fusion of Stereo and Motion This paper presents a new matching algorithm based on cooperative fusion of stereo and motion cues. In this algorithm, stereo disparity and image flow values are recovered from two successive pairs of stereo images by solving the stereo and motion corresponde

Love, A Treatise on Mathematical Theory of Elasticity (1944)(ISBN 0486601749).djvu 第三部分（共三部分）

Love, A Treatise on Mathematical Theory of Elasticity (1944)(ISBN 0486601749).djvu 第二部分（共三部分）

Love, A Treatise on Mathematical Theory of Elasticity (1944)(ISBN 0486601749) 第一部分（共三部分）

This paper describes an approach to real-time optical flow computation that combines the corner features and pyramid Lucas-Kanade. Corners instead of all the points in the image are taken into optical flow computation, which could reduce the amount of calculation to a large extend. The experiment has shown that using this optical flow algorithm to track targets is effective and could meet the requirements of real-time applications.

In this paper we present an approach to speed up the computation of sparse optical flow fields by means of integral images and provide implementation details. Proposing a modification of the Lucas-Kanade energy functional allows us to use integral images and thus to speed up the method notably while affecting only slightly the quality of the computed optical flow. The approach is combined with an efficient scanline algorithm to reduce the computation of integral images to those areas where there are features to be tracked. The proposed method can speed up current surveillance algorithms used for scene description and crowd analysis.

Medical Image Reconstruction: A Conceptual Tutorial" introduces the classical and modern image reconstruction technologies, such as two-dimensional (2D) parallel-beam and fan-beam imaging, three-dimensional (3D) parallel ray, parallel plane, and cone-beam imaging. This book presents both analytical and iterative methods of these technologies and their applications in X-ray CT (computed tomography), SPECT (single photon emission computed tomography), PET (positron emission tomography), and MRI (magnetic resonance imaging). Contemporary research results in exact region-of-interest (ROI) reconstruction with truncated projections, Katsevich's cone-beam filtered backprojection algorithm, and reconstruction with highly undersampled data with l0-minimization are also included.

Extraction and recognition of license plates of motorcycles and vehicles on highways

High Performance Implementation of License Plate Recognition in Image Sequences

Vs-star-- A visual interpretation system for visual surveillance

Robust fragments-based tracking with adaptive feature selection

Robust and automated unimodal histogram thresholding and potential applications

角点检测方法研究---根据实现方法不同可将角点检测方法分为两大类:基于边缘的角点检测方法与基于灰度变化的角点检测方法,并对现有的角点检测方法作了较为详细的分析与比较,指出角点检测技术的研究与发展方向.

图像融合中角点检测技术研究--图像融合中角点检测技术研究

Fast image region growing---Fast image region growing

Extracting Straight Lines---line detection edge detection

Corner Detection Algorithms for Digital Images in Last Three Decades

Application of Shape Analysis Techniques for the Classification of Vehicles

Solving the process of hysteresis without determining the optimal thresholds

second edition 2017, Oxford University Press

收录了超过100道机器学习算法工程师的面试题目和解答，本书将从特征工程、模型评估、降维等经典机器学习领域出发，构建一个算法工程师必-备的知识体系。其中大部分源于Hulu算法研究岗位的真实场景。

CLIP-Q: Deep Network Compression Learning by In-Parallel Pruning-Quantization，CVPR2018 code

Vehicle Detection and Tracking in Car Video Based on Motion Model--This work aims at real-time in-car video analysis to detect and track vehicles ahead for safety, auto-driving, and target tracing. This paper describes a comprehensive approach to localize target vehicles in video under various environmental conditions. The extracted geometry features from the video are projected onto a 1D profile continuously and are tracked constantly. We rely on temporal information of features and their motion behaviors for vehicle identification, which compensates for the complexity in recognizing vehicle shapes, colors, and types. We model the motion in the field of view probabilistically according to the scene characteristic and vehicle motion model. The Hidden Markov Model is used for separating target vehicles from background, and tracking them probabilistically. We have investigated videos of day and night on different types of roads, showing that our approach is robust and effective in dealing with changes in environment and illumination, and that real time processing becomes possible for vehicle borne cameras.

Projection and Least Square Fitting with Perpendicular Offsets based Vehicle License Plate Tilt Correction

Automatic video analysis from urban surveillance cameras is a fast-emerging field based on computer vision techniques. We present here a comprehensive review of the state-of-the-art computer vision for traffic video with a critical analysis and an outlook to future research directions. This field is of increasing relevance for intelligent transport systems (ITSs). The decreasing hardware cost and, therefore, the increasing de- ployment of cameras have opened a wide application field for video analytics. Several monitoring objectives such as congestion, traffic rule violation, and vehicle interaction can be targeted using cameras that were typically originally installed for human oper- ators. Systems for the detection and classification of vehicles on highways have successfully been using classical visual surveillance techniques such as background estimation and motion tracking for some time. The urban domain is more challenging with respect to traffic density, lower camera angles that lead to a high degree of occlusion, and the variety of road users. Methods from object categorization and 3-D modeling have inspired more advanced techniques to tackle these challenges. There is no commonly used data set or benchmark challenge, which makes the direct com- parison of the proposed algorithms difficult. In addition, evalu- ation under challenging weather conditions (e.g., rain, fog, and darkness) would be desirable but is rarely performed. Future work should be directed toward robust combined detectors and classifiers for all road users, with a focus on realistic conditions during evaluation.

This paper proposes two alternative formulations to reduce the high computational complexity of tensor voting, a robust perceptual grouping technique used to extract salient information from noisy data. The first scheme consists of numerical approximations of the votes, which have been derived from an in-depth analysis of the plate and ball voting processes. The second scheme simplifies the formulation while keeping the same perceptual meaning of the original tensor voting: The stick tensor voting and the stick component of the plate tensor voting must reinforce surfaceness, the plate components of both the plate and ball tensor voting must boost curveness, whereas junctionness must be strengthened by the ball component of the ball tensor voting. Two new parameters have been proposed for the second formulation in order to control the potentially conflictive influence of the stick component of the plate vote and the ball component of the ball vote. Results show that the proposed formulations can be used in applications where efficiency is an issue since they have a complexity of order O(1). Moreover, the second proposed formulation has been shown to be more appropriate than the original tensor voting for estimating saliencies by appropriately setting the two new parameters.

Marr and Hildreth's theory of LoG filtering with multiple scales has been extensively elaborated. One problem with LoG filtering is that it is very time-consuming, especially with a large size of filters. This paper presents a recursive convolution scheme for LoG filtering and a fast algorithm to extract zero-crossings. It has a constant computational complexity per pixel and is independent of the size of the filter. A line buffer is used to determine the locations of zero-crossings along with filtering hence avoiding the need for an additional convolution and extra memory units. Various images have been tested

Optimal filters for edge detection are usually developed in the continuous domain and then transposed by sampling to the discrete domain. Simpler filters are directly defined in the discrete domain. We define criteria to compare filter performances in the discrete domain. Canny has defined (1983, 1986) three criteria to derive the equation of an optimal filter for step edge detection: good detection, good localization, and low-responses multiplicity. These criteria seem to be good candidates for filter comparison. Unfortunately, they have been developed in the continuous domain, and their analytical expressions cannot be used in the discrete domain. We establish three criteria with the same meaning as Canny's.