Papers by Keyword: Corner Detection

Abstract: For mutual occlusion problem in multi-object tracking process, a novel tracking algorithm based on bilateral structure tensor corner detection is proposed, which can separate the objects correctly when they experience mutual occlusion. Firstly, it gains the information of each object corners. Secondly, when occlusion occurs, it makes use of K nearest neighbor algorithm combining with the nearest algorithm to classify the corners in occlusion region. Finally, the multi-object tracking algorithm is proposed. The experimental results show that the proposed method can separate the objects correctly and track the objects effectively, when they experience mutual occlusion, even the object changes its motion direction after occlusion.

Abstract: The inverse kinematics is very important in robotic application in the case of known object position. This paper explains about the inverse kinematics of right arm robot NAO based on the corner detection of a Chinese character image. The transformation coordinate system from image input of Chinese character to the robot NAO’s frame acceptable writing is needed. In order to control the right arm of robot NAO to write a Chinese character well, it is necessary to make a proper set of the RShoulderRoll angle q₁ and the RElbowRoll angle q₂, respectively. At first, the thinning processing, corner detection and depth-searching are employed to recognize a Chinese character. Due to the robot NAO just has three fingers, dust the position of the marker or pen-brush must be suitable of its. The experiment results achieve a better demonstration of recognition and handwriting for simple Chinese characters with less than 5 strokes.

Abstract: We present a camera calibration method based on vanishing point, that is, the vanishing points of two groups of parallel lines on the target plane are used to achieve camera calibration. A series of known positions points on target plane are used as the feature points, and the target images are recorded, the image coordinates of feature points are used to calculate the coordinates of vanishing point, then the matrix between feature points and camera is used to obtain internal parameters of camera. Experimental results show that the proposed calibration algorithm is correct, simple and convenient.

Abstract: According to the characteristics of Chinese characters image, we propose an improved corner detection method based on FAST algorithm and Harris algorithm to improve detection rate and shorten the running time for next feature extraction in this paper. The image of Chinese characters is detected for corners using FAST algorithm Firstly. Second, computing corner response function (CRF) of Harris algorithm, false corners are removed. The corners founded lastly are the endpoints of line segments, providing the length of line segments for shape feature extraction. The proposed method is compared with several corner detection methods over a number of images. Experimental results show that the proposed method shows better performance in terms of detection rate and running time.

Abstract: In computer vision system, the corners position need to be extracted from plane plate image. This paper presented a novel algorithm that improved the accuracy of corner detection from pixel to sub-pixel. The Canny operator was used to detect the corner edge pixels, and Gaussian filter was substituted by the bilateral filtering. It can remove noise and retain more slight edge information. Then, the corner edge pixels were transformed by Zernike moment and the sub-pixel edges of the corner were getted. Finally these sub-pixel edge points were linearly fitted and it was resulted in the corner coordinates of the intersection of two fitting straight lines. The experimental results show that the proposed method improves the corner detection precision to 0.1 pixel.

Abstract: Contour curve is an important shape feature for vehicle recognition and it is a hard work to extraction it from complex dynamic traffic video for in-vehicle detection system. Snake Model is used to automatically extract the object contour curve proposed by Kass et al, but it is inability for traffic objects. Presented here is a novel approach for extracting vehicle contour curve by combining stereo vision with Snake Model. In this paper, Stereo vision is first used to segment vehicle from traffic background, then Snake Model is adopted to obtain complete contour curve. In view of classical Snake model is easily affected by noise, here we propose a improved Snake model by combining corner detection technology with Distance Potential Snake Model. Moreover, a vehicle identification method based on contour curve is presented. The method presented here was tested on complex traffic scenes and the corresponding results prove the efficiency of our proposed method.

Abstract: Corner detection based on global and local curvature properties is an advanced method for detecting corners in images, which is a fundamental composition of many algorithms. However, we find that it is time-consuming for real-time applications and might detect wrong corners or lose some important corners. To alleviate these problems, we propose an improved curvature product corner detector with dynamic region of support based on Direct Curvature Scale Space (DCSS). Firstly, we use direct curvature scale space to reduce the complexity of computation instead of curvature scale space. Secondly, multi-scale curvature product with certain threshold is used to strengthen the corner detection. Finally, we check the angles of corner candidates in the dynamic region of support in order to eliminate falsely detected corners and use an adaptive curvature threshold to remove round corners from the initial list. The experimental results show that our proposed method improves the performance of corner detection both on accuracy and efficiency, and gain more stable corners at the same time.

Abstract: In order to improve accuracy and speed of image mosaic, an optimized algorithm of image mosaic based on corners is presented. Frequency phase correlation is used to estimate the overlapping area, in which improved Harris operator is used to extract corners. Then, rough and RANSAC accurate matching will be completed. Finally, image variance combined linear weight function is used to implemented image fusion and mosaic, and it presents quantitative analysis method about registration accuracy. Experimental results show that this algorithm exceeds existing ones at matching speed about 40%, and has a desired visual effect. The MAD and RMSE of corresponding points is within 1 pixel precision.

Abstract: There are a set of basic rules for stroke order in the traditional handwriting of Chinese characters, which may be listed as from left to right, from top to bottom, first inside then outside, and so on. Humanoid robot NAO is very famous in the world now, which has many sensors such as two cameras, infrared, IMU, bumper sensors and so forth. In this paper we explore how the humanoid robot NAO can write a Chinese character according to the traditional stroke order rules. As a matter of fact, the stroke order relation between any two strokes is very important in the whole stroke orders in order to lead a well writing. Therefore, the thinning algorithm is employed to propose an effective and useful method to determine the stroke order relation between any two strokes for Chinese character writing by robot NAO based on corner detection and depth-searching. Then the implementing algorithm is given for the determination of stroke order relation, and a series of experiments are carried out to validate its feasibility and effectiveness, the accuracy can achieve 90% so as to be a very satisfied result in the current stage.

Abstract: As a familiar mark in applications of computer vision, checkerboard can be used in camera calibration and pose measurement of dimensional object. When the checkerboard corners have been detected, for further calibration and measurement, the next necessary step is to rank these corners. This paper presents an automatic sort order of checkerboard corners based on convex hull Algorithm. As a result, the image dots of corners can be automatically matched with its corresponding physical dots without any artificial manipulation. Experimental results show the method is maneuverable and efficient with high stability.