Yingkui Du

A robust template tracking algorithm with weighted active drift correction

In this paper, we propose a novel algorithm for object template tracking and its drift correction. It can prevent the tracking drift effectively, and save the time of an additional correction tracking. In our algorithm, the total energy function consists of two terms: the tracking term and the drift correction term. We minimize the total energy function synchronously for template tracking and weighted active drift correction. The minimization of the active drift correction term is achieved by the inverse compositional algorithm with a weighted L2 norm, which is incorporated into traditional affine image alignment (AIA) algorithm. Its weights can be adaptively updated for each template. For diminishing the accumulative error in tracking, we design a new template update strategy that chooses a new template with the lowest matching error. Finally, we will present various experimental results that validate our algorithm. These results also show that our algorithm achieves better performance than the inverse compositional algorithm for drift correction.

The registration of UAV down-looking aerial images to satellite images with image entropy and edges

In this paper, we propose a novel and efficient image registration algorithm between high resolution satellite images and UAV down-looking aerial images. The algorithm is achieved by a composite deformable template matching. To overcome the limitations of environment changes and different sensors, and to remain image information, we fuse the image edge and entropy features as image representation. According to the altitude information of the UAV, we can get the scales of the down-looking aerial images relative to the satellite images. In the following, we perform an effective search strategy in the satellite images to find the best matching position. Different experimental results show that the proposed algorithm is effective and robust.

A double-side filter based power line recognition method for UAV vision system

Automatic power line recognition from cluttered background is an important and challenging task for a vision based unmanned aerial vehicles (UAVs) power line inspection system. In this paper, we propose a power line recognition method based on liner object enhancement and parallel lines constraint. A new double-side filter is proposed to enhance the power lines, and then the radon transform is used to find the parallel lines as the power line recognition results. Our experiments on real image data captured from UAV demonstrate our method is effective for automatic power line recognition.

Retinex theory‐based shadow detection and removal in single outdoor image

Purpose – Shadows, the common phenomena in most outdoor scenes, bring many problems in practical image processing. Shadow detection and removal, especial in uncalibrated outdoor image, is still a difficult problem. The purpose of this paper is to detect and to remove shadows in single outdoor image based on retinex theory.Design/methodology/approach – The shadow extraction algorithm originates from a simple idea that the human‐vision‐based retinex has the natural ability to enhance the shadow regions of an image no matter it is penumbrae or umbrae. Shadows are detected by comparing the retinex‐enhanced images with original images in the paper. The shadow removal algorithm in the paper deals with the shadow regions and non‐shadow regions in the images separately using the retinex enhancement algorithm. Through adding smooth light forcibly to shadow edges and introducing shadow edge masks, the authors reduce the effects of shadow edges in shadow removal processing.Findings – Some real single outdoor images ...

A novel color based object detection and localization algorithm

We propose a novel and robust color object detection and localization algorithm. Without a priori information about the number of objects, our method can detect all the objects with similar color feature in template. An improved histogram backprojection algorithm is used to find the object candidate regions. The weighted histogram intersection is used to verify the presence of objects. With the color feature in template, our method can detect and locate the objects accurately, get the number of objects, estimate their scales and orientations. Our experimental results on outdoor images obtained under different environments verify the effectiveness of our algorithm

Fast Ball Detection Method for Ping-Pong Playing Robots

Ping-pong ball detection plays an important role in the vision system of ping-pong playing robots. In the paper, we present a fast ball detection method for ping-pong playing robot. A reasonable searching region is proposed based on the ping-pong table and the prediction of the ball’s current position. Five key points are used to detect the ball in our method to meet the ping-pong ball’s high speed. Because we use points to the ping-pong ball instead of the ball’s region, the process of the detection can be finished during one scanning, which greatly increases the detecting speed. Experimental results verify the feasibility and the efficiency of our method.

A Real-Time Vision System for Telerobotic Manipulator with Large Time Delay

This paper presents a real-time vision system for telerobotic manipulator with large time delay. This system consists of a number of subsystems, including image acquisition system, image transmission system, object tracking system and pose estimation system. Furthermore, a pose estimation algorithm is proposed and developed to estimate the motion for tele-robotic manipulator by the 3D reconstruction uncertainty. In order to estimate the pose of the robotic manipulator, a new cooperative target is designed for tracking the manipulator. We validate our method through the experiments and comparison with the state-of-art methods. The performance of the vision system is validated in real tele-operation with a simulated large time delay.

Unsupervised progressive structure-from-motion for unordered images

A robust progressive structure from motion (PSFM) method is proposed for unordered images. Our method can reduce accumulative error efficiently during scene dense recovery and camera motion estimation. The whole unordered images are divided into two classes: key frames and non-key frames. For key frames, superior features are selected and tracked to initialize the parameters of PSFM reliably and reduce the accumulative errors. During the implementation of PSFM, several novel and efficient strategies are proposed to detect the parameter estimation errors and remove the outliers in feature tracking. A local on-demand scheme is proposed to dramatically reduce the computing cost of sparse bundle adjustment. The experimental results from different scenes show that our method is efficient and robust in scene dense recovery and camera motion estimation.

Adaptive rank transform for stereo matching

Window selection is the main challenge for local stereo matching methods based on the rank transform and it involves two aspects : the rank window selection and the match window selection. Most recent methods only focus on how to select the match window but pay little attention to the selection of the rank window. In this paper, we propose a novel matching method based on adaptive rank transform. Differing with the existing rank-based matching methods, the proposed method can deal with the rank and match window selection at the same time. The experimental results are evaluated on the Middlebury dataset as well as real images, showing that our method performs better than the recent rank-based stereo matching methods.

Efficient registration algorithm for UAV image sequence

This paper presents a fast and efficient image registration algorithm for UAV image sequence. The proposed algorithm consists of three main steps: feature extraction, feature point tracking, and homography matrix estimation. According to the comparison with different feature points, we choose the KLT feature and track them during the consecutive images. With the correct tracked feature points, the Total Least Squares (TLS) method is used to estimate homography matrix between two consecutive images, and then the multi-view constraint is used to refine the result and reduce the accumulative error in the image sequence. This method is robust and with less iteration times. Experiments on different image sequences indicate that our method has satisfactory image registration results with the average time 0.3s.