Inverse Perspective Transformation Research Articles

Lane Detection Based on Inverse Perspective Transformation and Kalman Filter

This paper proposes a novel algorithm for lane detection based on inverse perspective transformation and Kalman filter. A simple inverse perspective transformation method is presented to remove perspective effects and generate a top-view image. This method does not need to obtain the internal and external parameters of the camera. The Gaussian kernel function is used to convolute the image to highlight the lane lines, and then an iterative threshold method is used to segment the image. A searching method is applied in the top-view image obtained from the inverse perspective transformation to determine the lane points and their positions. Combining with feature voting mechanism, the detected lane points are fitted as a straight line. Kalman filter is then applied to optimize and track the lane lines and improve the detection robustness. The experimental results show that the proposed method works well in various road conditions and meet the real-time requirements.

Adaptive on-line calibration for around-view monitoring system using between-camera homography estimation

The around-view monitoring (AVM) system is one of the major applications of advanced driver assistance systems and intelligent transportation systems. We propose an on-line calibration method, which can compensate misalignments for AVM systems. Most AVM systems use fisheye undistortion, inverse perspective transformation, and geometrical registration methods. To perform these procedures, the parameters for each process must be known; the procedure by which the parameters are estimated is referred to as the initial calibration. However, when only using the initial calibration data, we cannot compensate misalignments, caused by changing equilibria of cars. Moreover, even small changes such as tire pressure levels, passenger weight, or road conditions can affect a car’s equilibrium. Therefore, to compensate for this misalignment, additional techniques are necessary, specifically an on-line calibration method. On-line calibration can recalculate homographies, which can correct any degree of misalignment using the unique features of ordinary parking lanes. To extract features from the parking lanes, this method uses corner detection and a pattern matching algorithm. From the extracted features, homographies are estimated using random sample consensus and parameter estimation. Finally, the misaligned epipolar geographies are compensated via the estimated homographies. Thus, the proposed method can render image planes parallel to the ground. This method does not require any designated patterns and can be used whenever cars are placed in a parking lot. The experimental results show the robustness and efficiency of the method.

비전 기반 실시간 원거리 2차원 바코드 검출 시스템

In this paper, we propose a real-time two-dimensional bar code detection system even at long range using a vision technique. We first perform short-range detection, and then long-range detection if the short-range detection is not successful. First, edge map generation, image binarization, and connect component labeling (CCL) are performed in order to select a region of interest (ROI). After interpolating the selected ROI using bilinear interpolation, a location symbol pattern is detected as the same as for short-range detection. Finally, the symbol pattern is arranged by applying inverse perspective transformation to localize bar codes. Experimental results demonstrate that the proposed system successfully detects bar codes at two or three times longer distance than existing ones even at indoor environment.

차선 변화벡터와 카디널 스플라인을 이용한 곡선 차선 검출방법

본 논문에서는 카메라 파라미터가 필요 없는 역 투시변환 영상에 차선 변화벡터와 카디널 스플라인을 이용하여 변화에 강인한 곡선 차선 검출 방법을 제안한다. 이 방법은 역 투시변환과 차선 필터의 전처리 과정이 적용된 영상의 시작 S 프레임과 그 다음 S+1 프레임에서 차선 후보 영역을 설정하여 차선 영역을 검출하고, 검출된 차선영역을 이용하여 차선 변화벡터를 계산한 결과를 가지고 이후의 프레임에서 차선이 위치할 지점을 예측한다. 이후에 예측된 차선 위치에서부터 스캔 영역을 설정하고 이 영역 내에서 새로운 차선 위치를 검출하며, 검출된 차선 위치를 이용해 차선 변화벡터를 갱신하고, 차선 영역 내의 제어 점들에 카디널 스플라인을 적용하여 차선을 검출한다. 제안하는 방법은 차선의 형태 변화에 강인한 곡선 차선 검출방법이지만 직선 차선에도 잘 적응됨을 보였으며 한 프레임을 처리하는 데 약 20ms 정도의 양호한 차선검출 속도를 보였다. The detection method of curves for the lanes which is powerful for the variation by utilizing the lane variation vector and cardinal spline on the inverse perspective transformation screen images which do not required the camera parameters are suggested in this paper. This method detects the lane area by setting the expected lane area in the s frame and next s+1 frame where the inverse perspective transformation and entire process of the lane filter are adapted, and expects the points of lane location in the next frames with the lane variation vector calculation from the detected lane areas. The scan area is set from the nextly expected lane position and new lane positions are detected within these areas, and the lane variation vectors are renewed with the detected lane position and the lanes are detected with application of cardinal spline for the control points inside the lane areas. The suggested method is a powerful method for curved lane detection, but it was adopted to the linear lanes too. It showed an excellent lane detection speed of about 20ms in processing a frame.

Robust ground plane region detection using multiple visual cues for obstacle avoidance of a mobile robot

SUMMARYThis paper presents a vision-based obstacle avoidance design using a monocular camera onboard a mobile robot. A novel image processing procedure is developed to estimate the distance between the robot and obstacles based-on inverse perspective transformation (IPT) in an image plane. A robust image processing solution is proposed to detect and segment a drivable ground area within the camera view. The proposed method integrates robust feature matching with adaptive color segmentation for plane estimation and tracking to cope with variations in illumination and camera view. After IPT and ground region segmentation, distance measurement results are obtained similar to those of a laser range finder for mobile robot obstacle avoidance and navigation. The merit of this algorithm is that the mobile robot can have the capacity of path finding and obstacle avoidance by using a single monocular camera. Practical experimental results on a wheeled mobile robot show that the proposed imaging system successfully obtains distances of surrounding objects for reactive navigation in an indoor environment.

A Method for Extracting QR Code from Complex Background Based on Morphology

In this paper we propose a new QR code extracting method based on morphology. Most of the time, locating Finder patterns is a significant part of QR code extraction. On the basis of traditional Finder Pattern detection method which checks whether certain areas meet 1:1:3:1:1 in both vertical and horizontal directions, we further refine the true Finder Patterns from several candidate areas through acreage proportion and gravity center detection, so as to eliminate interference from complex background. After image segmentation and getting the true finder patterns, other than the traditional method such as edge detection, we introduce the algorithm of region growth, along with choosing one seed pixel from obtained finder patterns to roughly figure out the QR code area. Eventually, by combining corner detection and inverse perspective transformation, we accomplish the extraction of QR code. Experiment results show that this method has robust correction capability from complex background and QR code deformation.

A monocular mobile robot reactive navigation approach based on the inverse perspective transformation

SUMMARYThis paper presents an approach to visual obstacle avoidance and reactive robot navigation for outdoor and indoor environments. The obstacle detection algorithm includes an image feature tracking procedure followed by a feature classification process based on the IPT (Inverse Perspective Transformation). The classifier discriminates obstacle points from ground points. Obstacle features permit to draw out the obstacle boundaries which are used to construct a local and qualitative polar occupancy grid, analogously to a visual sonar. The navigation task is completed with a robocentric localization algorithm to compute the robot pose by means of an EKF (Extended Kalman Filter). The filter integrates the world coordinates of the ground points and the robot position in its state vector. The visual pose estimation process is intended to correct possible drifts on the dead-reckoning data provided by the proprioceptive robot sensors. The experiments, conducted indoors and outdoors, illustrate the range of scenarios where our proposal has proved to be useful, and show, both qualitatively and quantitatively, the benefits it provides.

Concurrent visual navigation and localisation using inverse perspective transformation

A low-cost, simple and efficient approach to reactive visual navigation addressed to mobile robots, complemented with an accurate localisation algorithm' is presented. The system is mainly supported on the continuous detection and tracking of image features which are classified as obstacle or ground points using the inverse perspective transformation. While obstacle points are used for collision avoidance, ground points are used for localisation in a Rao-Blackwellised particle filter context. Results obtained from experiments conducted indoors and outdoors illustrate the wide range of scenarios where the technique can be successfully employed.

A New Approach to High-accuracy Road Orthophoto Mapping Based on Wavelet Transform

Existing orthophoto map based on satellite photography and aerial photography is not precise enough for road marking. This paper proposes a new approach to high-accuracy orthophoto mapping. The approach uses inverse perspective transformation to process the image information and generates the orthophoto fragment. The offline interpolation algorithm is used to process the location information. It processes the dead reckoning and the EKF location information, and uses the result to transform the fragments to the global coordinate system. At last it uses wavelet transform to divides the image to two frequency bands and uses weighted median algorithm to deal with them separately. The result of experiment shows that the map produced with this method has high accuracy.

블록 명암대비와 프로젝션에 기반한 2차원 바코드 검출 알고리즘

In an effort to increase the data capacity of one-dimensional symbology, 2D barcodes have been proposed a decade ago. In this paper, we present an effective 2D barcode detection algorithm from gray-level images, especially for the handheld 2D barcode recognition system. To locate the symbol inside the image, a criteria based on the block contrast is adopted, and a gray-scale projection with sub-pixel operation is utilized to segment the symbol precisely from the region of interest(ROI). Finally, the segmented ROI is normalized using the inverse perspective transformation for the following decoding processes. We also introduce the post-processing steps for decoding the QR-code. The proposed method ensures high performances under various lighting/printing conditions and strong perspective deformations. Experiments shows that our method is very robust and efficient in detecting the code area for the various types of 2D barcodes in real time.

로봇 착유시스템의 3차원 유두위치인식을 위한 스테레오비젼 시스템

A stereo vision system was developed for robot milking system (RMS) using two monochromatic cameras. An algorithm for inverse perspective transformation was developed for the 3-D information acquisition of all teats. To verify performance of the algorithm in the stereo vision system, indoor tests were carried out using a test-board and model teats. A real cow and a model cow were used to measure distance errors. The maximum distance errors of test-board, model teats and real teats were 0.5 mm, 4.9 mm and 6 mm, respectively. The average distance errors of model teats and real teats were 2.9 mm and 4.43 mm, respectively. Therefore, it was concluded that this algorithm was sufficient for the RMS to be applied.

젖소의 개체인식 및 형상 정보화를 위한 컴퓨터 시각 시스템 개발(II) - 스테레오 영상을 이용한 체위 분석 -

The analysis of cow body parameters is important to provide some useful information fur cow management and cow evaluation. Present methods give many stresses to cows because they are invasive and constrain cow postures during measurement of body parameters. This study was conducted to develop the stereo vision system fur non-invasive analysis of cow body features. Body feature parameters of 16 heads at two farms(A, B) were measured using scales and nineteen stereo images of them with walking postures were captured under outdoor illumination. In this study, the camera calibration and inverse perspective transformation technique was established fer the stereo vision system. Two calibration results were presented for farm A and fm B, respectively because setup distances from camera to cow were 510 cm at farm A and 630cm at farm B. Calibration error values fer the stereo vision system were within 2 cm for farm A and less than 4.9 cm for farm B. Eleven feature points of cow body were extracted on stereo images interactively and five assistant points were determined by computer program. 3D world coordinates for these 15 points were calculated by computer program and also used for calculation of cow body parameters such as withers height. pelvic arch height. body length. slope body length. chest depth and chest width. Measured errors for body parameters were less than 10% for most cows. For a few cow. measured errors for slope body length and chest width were more than 10% due to searching errors fer their feature points at inside-body positions. Equation for chest girth estimated by chest depth and chest width was presented. Maximum of estimated error fur chest girth was within 10% of real values and mean value of estimated error was 8.2cm. The analysis of cow body parameters using stereo vision system were successful although body shape on the binocular stereo image was distorted due to cow movements.

Kalman Filter Approach for Lane Extraction and Following

In this paper, we focus on the use of Kalman filter approach to lane extraction and following. We assume a structured environment where a mobile vehicle equipped with a camera sensor, situated at a fixed distance from the vehicle frame, is navigating. A quadratic model of the road is considered. This enables the state vector of the filter to be coincided with the three parameters pertaining to a second-order polynom. The determination of the state model is carried out considering either a pure translation of the (i+1)th frame with respect to the ith frame attached respectively to the images at time ti and ti+1, or a combination of both translation and rotation. While the measurement model comes down to the camera images obtained using the inverse perspective transformation, which due to the road model, permits a straightforward link between the x–y co-ordinates of images and the state vector parameters of the road. The performance of these two state models are compared with a purely measurement approach where least squares methodology is performed regardless the state model. The estimates of the filter are used by the vehicle in order to update its own knowledge about the environment and to accomplish the task of “road following”. Furthermore, this permits the vision system to encompass into a more general structure of Integrated Supervisory Control Systems (ISCS), where multiple functionalities are allowed.

Building a local floor map by use of ultrasonic and omni-directional vision sensors

For robot navigation, it is important to have a local map which gives the free space around the robot. Various sensing and fusion methods have been developed to detect the free space. In this paper we propose a new fusion approach which uses the ultrasonic sensor aided by an omni-directional vision sensor to give a grid-based free space around the robot. By use of the ultrasonic sensor, the robot can obtain conservative range information based on our nearby range filtering method. The filtering can give a more reliable result considering the sensor's problem of specular reflection. Also, by use of the special omni-directional vision sensor we developed, color and edge information can be obtained in a single picture and mapped to the ground plane by the inverse perspective transformation. Thus the range, color and edge information can all be expressed on a metric grid-based representation which forms the basis of our fusion processing. From the filtered range information in this metric representation, we obtain an initial safety index for each grid cell around the robot. From the color and edge information, uniform color regions and connectivity information are extracted and used to revise the safety index. The grid cells whose final safety index is above some given threshold will then constitute the free space around the robot. Results in an indoor cluttered environment are given which show the usefulness of our proposed sensor fusion approach.

Uncalibrated obstacle detection using normal flow

This paper addresses the problem of obstacle detection for mobile robots. The visual information provided by a single on-board camera is used as input. We assume that the robot is moving on a planar pavement, and any point lying outside this plane is treated as an obstacle. We address the problem of obstacle detection by exploiting the geometric arrangement between the robot, the camera, and the scene. During an initialization stage, we estimate an inverse perspective transformation that maps the image plane onto the horizontal plane. During normal operation, the normal flow is computed and inversely projected onto the horizontal plane. This simplifies the resultant flow pattern, and fast tests can be used to detect obstacles. A salient feature of our method is that only the normal flow information, or first order time-and-space image derivatives, is used, and thus we cope with the aperture problem. Another important issue is that, contrasting with other methods, the vehicle motion and intrinsic and extrinsic parameters of the camera need not be known or calibrated. Both translational and rotational motion can be dealt with. We present motion estimation results on synthetic and real-image data. A real-time version implemented on a mobile robot, is described.

Recognition of planar shapes from perspective images using contour-based invariants

This paper presents a new method of recognizing a two-dimensional shape from a single perspective projection image taken from an unknown (three-dimensional) viewpoint. The method is based on quadratic approximations to the effect of the slant of an inverse perspective transformation on angles and lengths. These approximations allow us to define contour-based properties that are invariant under perspective transformation. The method can be used to recognize partially occluded shapes, as well as shapes that are not exactly related by perspective transformations. When a shape is recognized, the method also provides estimates of its tilt and slant.

Motion stereo for navigation of autonomous vehicles in man-made environments

In order to navigate, vehicles need to deduce their movements in given environments. We describe a method for wide-angle motion stereo to aid such navigations. Our approach uses vertices of objects observed in a scene as features to be matched in the two images. The matching is done in the 3-d world with the guidance of a base line which can be seen in both images. The candidate vertices for matching are obtained from analyzing (1) the edge maps of input images using a junction type table and (2) the range information of the segments which form the vertices. Range information is obtained by inverse perspective transformation. Some experimental results are shown.

Using perspective transformations in scene analysis

This paper briefly reviews in a self-contained manner the perspective transformation governing 2-D images taken of a 3-D world. Using coordinates of points, direction cosines of lines on the image, or 3-D models of objects, relationships are developed that permit determination of the perspective transformation parameters in closed form, as well as the 3-D coordinates of objects. Contents. I, Introduction; I.1, 3-D Scene Analysis. II, The Perspective Transformation in 2-D. III, The Perspective Transformation in 3-D. IV, Properties of the 3-D Perspective Transformation; IV.1, Lines to Lines; IV.2, Vanishing Points. V, Image Coordinates to 3-D Coordinates; V.1, Camera Geometry; V.2, Vanishing Point to Camera Parameters; V.3, The Inverse Projective Transformation; V.4, Using the Inverse Perspective Transformation; V.5, Using the Perspective Transformation; V.6, Models. VI, Example; VI.1, Example 1; VI.2, Example 2. Acknowledgement. References.