Showing papers on "Aerial image published in 2010"

Learning to detect roads in high-resolution aerial images

Abstract: Reliably extracting information from aerial imagery is a difficult problem with many practical applications. One specific case of this problem is the task of automatically detecting roads. This task is a difficult vision problem because of occlusions, shadows, and a wide variety of non-road objects. Despite 30 years of work on automatic road detection, no automatic or semi-automatic road detection system is currently on the market and no published method has been shown to work reliably on large datasets of urban imagery. We propose detecting roads using a neural network with millions of trainable weights which looks at a much larger context than was used in previous attempts at learning the task. The network is trained on massive amounts of data using a consumer GPU. We demonstrate that predictive performance can be substantially improved by initializing the feature detectors using recently developed unsupervised learning methods as well as by taking advantage of the local spatial coherence of the output labels.We show that our method works reliably on two challenging urban datasets that are an order of magnitude larger than what was used to evaluate previous approaches.

Human identity recognition in aerial images

Abstract: Human identity recognition is an important yet under-addressed problem. Previous methods were strictly limited to high quality photographs, where the principal techniques heavily rely on body details such as face detection. In this paper, we propose an algorithm to address the novel problem of human identity recognition over a set of unordered low quality aerial images. Assuming a user was able to manually locate a target in some images of the set, we find the target in each other query image by implementing a weighted voter-candidate formulation. In the framework, every manually located target is a voter, and the set of humans in a query image are candidates. In order to locate the target, we detect and align blobs of voters and candidates. Consequently, we use PageRank to extract distinguishing regions, and then match multiple regions of a voter to multiple regions of a candidate using Earth Mover Distance (EMD). This generates a robust similarity measure between every voter-candidate pair. Finally, we identify the candidate with the highest weighted vote as the target. We tested our technique over several aerial image sets that we collected, along with publicly available sets, and have obtained promising results.

Automatic Detection of Buildings and Changes in Buildings for Updating of Maps

Abstract: There is currently high interest in developing automated methods to assist the updating of map databases. This study presents methods for automatic detection of buildings and changes in buildings from airborne laser scanner and digital aerial image data and shows the potential usefulness of the methods with thorough experiments in a 5 km2 suburban study area. 96% of buildings larger than 60 m2 were correctly detected in the building detection. The completeness and correctness of the change detection for buildings larger than 60 m2 were about 85% (including five classes). Most of the errors occurred in small or otherwise problematic buildings.

Real-time aerial image mosaicing

Abstract: This paper describes a scheme for seamlessly stitching together images captured from an aerial platform, in real-time, in order to provide an operator with a larger field-of-view. Both recent images, and images from earlier in a flight are used. To obtain real-time performance several of the latest computer vision techniques are applied: firstly the Bag-of-Words image representation allows overlapping images to be found efficiently, and provides cheap wide-baseline correspondences between them. Secondly the BaySAC robust estimation framework allows images to be registered efficiently from a prior motion model combined with large numbers of potential matches between cheap image patch descriptors. Thirdly an efficient seam-placement algorithm allows the rendering of a visually attractive mosaic. Results are presented on a sequence of high-resolution images captured from a microlight.

A Hierarchical and Contextual Model for Aerial Image Parsing

Abstract: In this paper we present a hierarchical and contextual model for aerial image understanding. Our model organizes objects (cars, roofs, roads, trees, parking lots) in aerial scenes into hierarchical groups whose appearances and configurations are determined by statistical constraints (e.g. relative position, relative scale, etc.). Our hierarchy is a non-recursive grammar for objects in aerial images comprised of layers of nodes that can each decompose into a number of different configurations. This allows us to generate and recognize a vast number of scenes with relatively few rules. We present a minimax entropy framework for learning the statistical constraints between objects and show that this learned context allows us to rule out unlikely scene configurations and hallucinate undetected objects during inference. A similar algorithm was proposed for texture synthesis (Zhu et al. in Int. J. Comput. Vis. 2:107---126, 1998) but didn't incorporate hierarchical information. We use a range of different bottom-up detectors (AdaBoost, TextonBoost, Compositional Boosting (Freund and Schapire in J. Comput. Syst. Sci. 55, 1997; Shotton et al. in Proceedings of the European Conference on Computer Vision, pp. 1---15, 2006; Wu et al. in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 1---8, 2007)) to propose locations of objects in new aerial images and employ a cluster sampling algorithm (C4 (Porway and Zhu, 2009)) to choose the subset of detections that best explains the image according to our learned prior model. The C4 algorithm can quickly and efficiently switch between alternate competing sub-solutions, for example whether an image patch is better explained by a parking lot with cars or by a building with vents. We also show that our model can predict the locations of objects our detectors missed. We conclude by presenting parsed aerial images and experimental results showing that our cluster sampling and top-down prediction algorithms use the learned contextual cues from our model to improve detection results over traditional bottom-up detectors alone.

Moving Objects Detection and Tracking Framework for UAV-based Surveillance

Abstract: Automated motion detection and tracking of ground moving objects using aerial platforms is challenging due to the small object size in comparison with objects such as buildings, as well as the fact that flying cameras can undergo rapid translations and rotations. As such, our objectives are to develop a system for gathering useful information from aerial images by mapping visited areas through image mosaic king and to detect moving objects in the captured video. To do so, the Moving Objects Detection and Tracking (MODAT) framework has been developed to facilitate the application and combination of various relevant computer vision and image processing techniques in order to achieve our objectives.

Vehicle ego-localization by matching in-vehicle camera images to an aerial image

Abstract: Obtaining an accurate vehicle position is important for intelligent vehicles in supporting driver safety and comfort. This paper proposes an accurate ego-localization method by matching in-vehicle camera images to an aerial image. There are two major problems in performing an accurate matching: (1) image difference between the aerial image and the in-vehicle camera image due to view-point and illumination conditions, and (2) occlusions in the in-vehicle camera image. To solve the first problem, we use the SURF image descriptor, which achieves robust feature-point matching for the various image differences. Additionally, we extract appropriate feature-points from each road-marking region on the road plane in both images. For the second problem, we utilize sequential multiple in-vehicle camera frames in the matching. The experimental results demonstrate that the proposed method improves both ego-localization accuracy and stability.

Impacts of cost functions on inverse lithography patterning.

Abstract: For advanced CMOS processes, inverse lithography promises better patterning fidelity than conventional mask correction techniques due to a more complete exploration of the solution space. However, the success of inverse lithography relies highly on customized cost functions whose design and know-how have rarely been discussed. In this paper, we investigate the impacts of various objective functions and their superposition for inverse lithography patterning using a generic gradient descent approach. We investigate the most commonly used objective functions, which are the resist and aerial images, and also present a derivation for the aerial image contrast. We then discuss the resulting pattern fidelity and final mask characteristics for simple layouts with a single isolated contact and two nested contacts. We show that a cost function composed of a dominant resist-image component and a minor aerial-image or image-contrast component can achieve a good mask correction and contour targets when using inverse lithography patterning.

Detection of building changes from aerial images and light detection and ranging (LIDAR) data

Abstract: Building models are built to provide three-dimensional (3-D) spatial information, which is needed in a variety of applications including city planning, construction management, location-based services of urban infrastructures, and the like. However, 3-D building models have to be updated on a timely manner to meet the changing demand. Rather than reconstructing building models for the entire area, it would be more convenient and effective to only update parts of the areas where there were changes. This paper aims at developing a new method, namely double-threshold strategy, to find such changes within 3-D building models in the region of interest with the aid of light detection and ranging (LIDAR) data. The proposed modeling scheme comprises three steps, namely, data pre-processing, change detection in building areas, and validation. In the first step for data pre-processing, data registration was carried out based on multi-source data. The second step for data pre-processing requires using the triangulation of an irregular network of data points collected by Light Detection And Ranging (LIDAR), focusing on those locations containing walls or other above-ground objects that were ever removed. Then, change detection in the building models can be made possible for finding differences in height by comparing the LIDAR point measurements and the estimates of the building models. The results may be further refined using spectral and feature information collected from aerial imagery. A double-threshold strategy was applied to cope with the highly sensitive thresholding often encountered when using the rule-based approach. Finally, ground truth data were used for model validation. Research findings clearly indicate that the double-threshold strategy improves the overall accuracy from 93.1% to 95.9%.

A Method to Extract Insulator Image From Aerial Image of Helicopter Patrol

Abstract: A new algorithm to extract insulator images from aerial image acquired by intelligent patrol of helicopter is proposed.Firstly,this algorithm converts the color image of glass insulators with high resolution,which is acquired by helicopter patrol,from RGB color space into HSI color space;then using the maximum entropy threshold method based on genetic algorithm,the continue image segmentation is applied to S component in HSI space;and then the noise in the segmented image is filtered by doubly structured cascaded filters;finally,by means of connected components labeling operation the contour of insulator string is marked out from aerial image with complicated background.Results of calculation example show that the proposed algorithm can extract integral insulator image from aerial image with complicated background,and it is practicable.

Fast pattern matching

Abstract: Disclosed are methods, systems, and articles of manufacture for using pattern matching with an integrated circuit layout including recognizing shapes within the IC layout, identifying features for the shapes, and extracting situations for the respective features. The method may further include simulating the situations to determine a set of situations for modification based on an OPC requirement, modifying the set of situations to improve satisfaction of the OPC requirement, and reintegrating the modified set of situations into the IC layout. The method may also include simulating a subset of the extracted situations to determine aerial images of the subset, and tiling the subset of situations to form a larger aerial image. The method may also include removing overlap from a window based on the situations extracted for the window, calculating a density for each of the situations, and calculating a density for the window based on the density.

A hybrid moving object detection method for aerial images

Abstract: Compared to stationary surveillance cameras, moving object detection on the camera carried by airborne vehicle is more difficult because of the relatively dynamic background. In this study, we present a hybrid method for moving object detection in aerial videos. We compensated the ego motion of airborne vehicle by feature-point based image alignment on consecutive frames, and then applied an accumulative frame differencing method to detect the pixels with motion. Meanwhile, the current frame was divided into homogenous regions by image segmentation, and some of them were selected as candidates of moving objects by prior rules. The motion pixels and the moving object candidates were then fused by a morphing-based approach, obtaining position and shape of moving objects. Moreover, a Kalman-filter tracker was adopted to not only give a consistent label on each detected moving object and but also reject false alarms. The proposed method was evaluated on the videos captured on an airborne vehicle at different altitude. Experimental results revealed that the proposed hybrid method has better performance than both frame-difference and optical-flow based approaches.

Real-time image processing for road traffic data extraction from aerial images

Abstract: A world with growing individual traffic requires sufficient solutions for traffic monitoring and guidance. The actual ground based approaches for traffic data collection may be barely sufficient for everyday life, but they will fail in case of disasters and mass events. Therefore, a road traffic monitoring solution based on an airborne wide area camera system has been currently developed by DLR. Here, we present a new image processing chain for real-time traffic data extraction from high resolution aerial image sequences with automatic methods. This processing chain is applied in a computer network as part of an operational sensor system for traffic monitoring onboard a DLR aircraft. It is capable of processing aerial images obtained with a frame rate of up to 3 Hz. The footprint area of the three viewing directions of an image exposure with three cameras is 4 x 1 km at a resolution of 20 cm (recorded at a flight height of 1500 m). The processing chain consists of a module for data readout from the cameras and for the synchronization of the images with the GPS/IMU navigation data (used for direct georeferencing) and a module for orthorectification of the images. Traffic data is extracted by a further module based on a priori knowledge from a road database of the approximate location of road axes in the georeferenced and orthorectified images. Vehicle detection is performed by a combination of Adaboost using Haar-like features for pixel wise classification and subsequent clustering by Support Vector Machine based on a set of statistical features of the classified pixel. In order to obtain velocities, vehicle tracking is applied to consecutive images after performing vehicle detection on the first image of the burst. This is done by template matching along a search space aligned to road axes based on normalized cross correlation in RGB color space. With this processing chain we are able to obtain accurate traffic data with completeness and correctness both higher than 80% at high actuality for varying and complex image scenes. The proposed processing chain is evaluated on a huge number of images including inner city scenes of Cologne and Munich, demonstrating the robustness of our work in operational use.

Horizon detection in foggy aerial image

Abstract: Vision-based automatically landing is important for micro Unmanned Aerial Vehicles (UAVs). Horizon is a very useful clue. Most of the existing solutions for the problem can get accurate results in clear weather. However, for some images shoot in extreme environmental conditions like foggy or cloudy sky these methods are difficult in identifying the horizon correctly. In this paper, we propose a robust, vision-based horizon detection algorithm fit for this condition. The algorithm we put forward is based on a dark channel prior, which describes the depth of haze naturally. The horizon can be easily determined in dark channel property space. We then verify our vision-based horizon detection algorithm with real flying data. The results indicate that the algorithm is robust to heavy foggy weather conditions. This algorithm can also be useful in synthetic vision system.

Building Detection in a Single Remotely Sensed Image with a Point Process of Rectangles

Abstract: In this paper we introduce a probabilistic approach of building extraction in remotely sensed images. To cope with data heterogeneity we construct a flexible hierarchical framework which can create various building appearance models from different elementary feature based modules. A global optimization process attempts to find the optimal configuration of buildings, considering simultaneously the observed data, prior knowledge, and interactions between the neighboring building parts. The proposed method is evaluated on various aerial image sets containing more than 500 buildings, and the results are matched against two state-of-the-art techniques.

Techniques for atmospheric and solar correction of aerial images

Abstract: Techniques for atmospheric and solar correction of aerial images are described. An apparatus may comprise an atmospheric and solar component arranged for execution by a logic device and operative to correct solar and atmosphere artifacts from an aerial image. The atmospheric and solar component may comprise an image information component operative to generate an image record for each aerial image of a group of aerial images, the image record comprising statistical information and image context information for each aerial image, a filter generation component operative to generate an atmospheric filter and a solar filter from the statistical information and the image context information stored in the image records, and an image correction component operative to correct atmospheric and solar artifacts from the aerial image using the respective atmospheric filter and solar filter. Other embodiments are described and claimed.

Tree detection form aerial imagery

Abstract: Methods and systems for detecting trees in aerial imagery are provided. A method for detecting trees in aerial imagery may include training a pixel-level classifier to assign a tree or non-tree label to each pixel in an aerial image. The method may further include segmenting tree and non-tree regions of the labeled pixels. The method may also include locating individual tree crowns in the segmented tree regions using one or more tree templates. A system for detecting trees in aerial imagery may include a trainer, a segmenter and a tree locator.

Automatic registration of oblique aerial images with cadastral maps

Abstract: In recent years, oblique aerial images of urban regions have become increasingly popular for 3D city modeling, texturing, and various cadastral applications. In contrast to images taken vertically to the ground, they provide information on building heights, appearance of facades, and terrain elevation. Despite their widespread availability for many cities, the processing pipeline for oblique images is not fully automatic yet. Especially the process of precisely registering oblique images with map vector data can be a tedious manual process. We address this problem with a registration approach for oblique aerial images that is fully automatic and robust against discrepancies between map and image data. As input, it merely requires a cadastral map and an arbitrary number of oblique images. Besides rough initial registrations usually available from GPS/INS measurements, no further information is required, in particular no information about the terrain elevation.

Automatic vehicle detection in aerial image sequences of urban areas using 3d hog features

Abstract: With the development of low cost aerial optical sensors having a spatial resolution in the range of few centimetres, the traffic monitoring by plane receives a new boost. The gained traffic data are very useful in various fields. Near real-time applications in the case of traffic management of mass events or catastrophes and non time critical applications in the wide field of general transport planning are considerable. A major processing step for automatically provided traffic data is the automatic vehicle detection. In this paper we present a new processing chain to improve this task. First achievement is limiting the search space for the detector by applying a fast and simple pre-processing algorithm. Second achievement is generating a reliable detector. This is done by the use of HoG features (Histogram of Oriented Gradients) and their appliance on two consecutive images. A smart selection of this features and their combination is done by the Real AdaBoost (Adaptive Boosting) algorithm. Our dataset consists of images from the 3K camera system acquired over the city of Munich, Germany. First results show a high detection rate and good reliability.

Tone mapping of very large aerial image mosaic

Abstract: A method for tone mapping a high dynamic range image of a large terrestrial area into a lower dynamic range image uses a globally aware, locally adaptive approach whereby local tonal balancing parameter values are derived from known tone mapping parameters for a local 3×3 matrix of image tiles and used in turn to derive a local sigmoid transfer function for pixels in the tile in the middle of the matrix. A global sigmoid transfer function is derived based on values of the tone mapping parameters applicable to the entire image. A lower dynamic range image pixel will have a local tone mapped value and a globally tone mapped value, which are combined by giving each a weighted value to provide a final low dynamitic range pixel value.

Three-Dimensional Semiautomatic Road Extraction From a High-Resolution Aerial Image by Dynamic-Programming Optimization in the Object Space

Abstract: This letter proposes a method for 3-D semiautomatic road extraction from a single image by dynamic programming (DP), which requires a form of relief representation such as digital terrain model. Unlike traditional DP methodologies, this method relies on the DP algorithm to carry out the optimization process in the object space rather than in the image space. Road features are traced in the object space, implying the need for a rigorous mathematical model between image- and object-space points. The operator must measure a few seed points in the image space to coarsely describe the roads, which are then transformed into object space to initialize the DP optimization process. The results showed that the method usually works properly, even in the presence of anomalies along roads.

Extreme ultraviolet laser-based table-top aerial image metrology of lithographic masks

Abstract: We have realized the first demonstration of a table-top aerial imaging microscope capable of characterizing pattern and defect printability in extreme ultraviolet lithography masks. The microscope combines the output of a 13.2 nm wavelength, table-top, plasma-based, EUV laser with zone plate optics to mimic the imaging conditions of an EUV lithographic stepper. We have characterized the illumination of the system and performed line-edge roughness measurements on an EUVL mask. The results open a path for the development of a compact aerial imaging microscope for high-volume manufacturing.

Magnifying imaging lens for use in aerial image metrology system for e.g. simulation of effects of characteristics of lithography masks used for manufacturing semiconductor elements, has image plane representing lens field plane

Abstract: The lens (1) has a mirror group with two mirrors (M1, M2) and a diffractive optical element (11) i.e. zone plate, for imaging an object field (2) in an object plane (3) in an image field (4) in an image plane (5). The image plane represents a lens field plane after the object field. A spatial distance (A) between one of the mirrors, which is arranged spatially nearest to the image field, and the image field is smaller than a spatial distance (B) between the element and the image field. The object field and the image field are arranged in center relative to an optical axis (oA) of the lens. An independent claim is also included for a metrology system for inspecting objects.

Extreme ultraviolet mask substrate surface roughness effects on lithographic patterning

Abstract: In extreme ultraviolet lithography exposure systems, mask substrate roughness-induced scatter contributes to line edge roughness (LER) at the image plane. In this article, the impact of mask substrate roughness on image plane speckle is explicitly evaluated. A programed roughness mask was used to study the correlation between mask roughness metrics and wafer plane aerial image inspection. The authors find that the roughness measurements by the top surface topography profile do not provide complete information on the scatter related speckle that leads to LER at the image plane. They suggest at-wavelength characterization by imaging and/or scatter measurements into different frequencies as an alternative for a more comprehensive metrology of the mask substrate/multilayer roughness effects.

Photo-Mask Acceptance Technique

Abstract: A technique for calculating a second aerial image associated with a photo-mask that can be used to determine whether or not the photo-mask (which may include defects) is acceptable for use in a photolithographic process is described. In particular, using a first aerial image produced by the photo-mask when illuminated using a source pattern and an inspection image of the photo-mask, a mask pattern corresponding to the photo-mask is determined. For example, the first aerial image may be obtained using an aerial image measurement system, and the inspection image may be a critical-dimension scanning-electron-microscope image of the photo-mask. This image, which has a higher resolution than the first aerial image, may indicate spatial-variations of a magnitude of the transmittance of the photo-mask. Then, the second aerial image may be calculated based on the determined mask pattern using a different source pattern than the source pattern.

People tracking and trajectory interpretation in aerial image sequences

Abstract: Monitoring the behavior of people in complex environments has gained much attention over the past years. Most of the current approaches rely on video cameras mounted on buildings or pylons and people are detected and tracked in these video streams. The presented approach is intended to complement this work. The monitoring of people is based on aerial image sequences derived with camera systems mounted on aircrafts, helicopters or airships. This imagery is characterized by a very large coverage providing the opportunity to analyze the distribution of people over a large field of view. The approach shows first results on automatic detection and tracking of people from image sequences. In addition, the derived trajectories of the people are automatically interpreted to reason about the behavior and to detect exceptional events.

Relationship between Normalized Image Log Slope and Chemical Gradient in Chemically Amplified Extreme Ultraviolet Resists

Abstract: In lithography, normalized image log slope (NILS) is an important metric that describes the quality of an aerial image of incident photons. The aerial image is converted to a latent image through lithographic processes in the resist. The quality of the latent image correlates with line edge roughness (LER). Chemical gradient is also an important metric that describes the quality of a latent image. In this study, we investigated the relationship between NILS and chemical gradient in chemically amplified resists for extreme ultraviolet lithography. In particular, the effect of effective reaction radius for catalytic chain reactions on the relationship between NILS and chemical gradient was clarified.

Method for emulation of a photolithographic process and mask inspection microscope for performing the method

Abstract: In mask inspection, the defects that are of interest are primarily those that will also show up on wafer exposure. The aerial images generated in the resist and by emulation should be as identical as possible. This also applies to methods in which an overall structure that is divided into at least two substructures on at least two masks. A system and a method are provided for emulating a photolithographic process for generating on a wafer an overall structure that is divided into at least two substructures on at least two masks. The method includes generating aerial images of the at least two substructures, at least one of the aerial images being captured with a mask inspection microscope; correcting, by using a processing unit, errors in the at least one aerial image captured with a mask inspection microscope; and overlaying the aerial images of the at least two substructures to form an overall aerial image with the overall structure.