Current Issue Cover


摘 要
目的 摄像机旋转扫描条件下的动目标检测研究中,传统的线性模型无法解决摄像机旋转扫描运动带来的图像间非线性变换问题,导致图像补偿不准确,在动目标检测时将引起较大误差,造成动目标虚假检测。为解决这一问题,提出了一种面阵摄像机旋转扫描条件下的图像补偿方法,其特点是能够同时实现背景运动补偿和图像非线性变换补偿,从而实现动目标的快速可靠检测。方法 该方法首先进行图像匹配,然后建立摄像机旋转扫描非线性模型,通过参数空间变换将其转化为线性求解问题,采用Hough变换实现该方程参数的快速鲁棒估计。解决摄像机旋转扫描条件下获取的图像间非线性变换问题,从而实现图像准确补偿。在此基础上,可以利用帧间差分等方法检测出运动目标。结果 实验结果表明,在摄像机旋转扫描条件下,本文方法能够同时实现图像间的背景运动补偿和非线性变换补偿,可以去除大部分由于立体视差效应(parallax effects)产生的匹配错误。并且在实验中,本文的方法处理速度可以达到50frames/s,满足实时性要求。结论 在面阵摄像机旋转扫描的条件下,相比于传统的基于线性模型的图像补偿方法,本文的方法能够快速、准确地在背景补偿的基础上同时解决图像间非线性变换问题,从而更好的提取出运动目标,具有一定的实用价值。
Image compensation for object detection under rotating camera

Zhai Dingding,Wang Qi,yangyan,Wang Fan,Hu Xiaopeng(Dalian University of Technology;Liaoning Normal University)

Objective In the field of moving object detection, detecting on fixed cameras has been gradually matured. In many practical applications, camera motion such as rotating-scan is required to increase the monitoring range and achieve gaze monitoring. Compared wth the moving object detection under fixed camera conditions, camera’s motion makes moving object detection more difficult. Because of image transfomation caused by camera’s motion, image compensation is needed to eliminate their effect. However, the traditional linear model can not solve the nonlinear transform that is generated by the rotating-scan movement of cameras. Under the condition of camera rotating-scan, the key step of image compensation is to find an accurate motion model to describe image transformations between image frames, including rotation, translation and scaling. The existing methods is not able to meet the application requirements in terms of calculation time and accuracy simultaneously. In order to solve this problem, a robust image compensation method under the condition of camera rotating-scan is proposed, which can achieve background motion compensation and image nonlinear transform compensation simultaneously. Method Our method contains four steps to achieve the goal of image compensation for camera rotating-scan. Firstly, corresponding point pairs are obtained through image matching. Feature points in the current frame are extracted by FAST corner detection method, and then matched with those in the previous frame. Next the global displacement of the background is computed through the matching points. On this basis, Kalman filter updates its state and predict the global displacement of the next frame and the positions of the current feature points appearing in the next image. As a consequence, the feature points in the next frame matched with the current feature point will be searched in the estimated image area. As the matched image area is reduced, feature matching accuracy can be improved. Secondly, a global transformation model between adjacent frames is established, According to the analyzation of the camera imaging mechanism for rotating-scan, a nonlinear motion model is proposed. Based on the nonlinear motion model, a camera equation is established, which is transformed into a linear problem further by the parameter space conversion. Thirdly, Hough transform is utilized to estimate the parameters of the global motion model by the matched point pairs. The global motion model is then mapped into the image to obtain the coordinate transformation relationship between adjacent images. Through the coordinate transformation relationship, the image is normalized to a unified coordinate system. This step leads to the implementation of background motion compensation and nonlinear transform compensation. Finally, foreground objects are segmented from the image. The block based inter-frame difference method is used to detect moving objects. In order to extract the foreground objects completely, the mathematical morphological opening is operated to eliminate isolated pixel points and small line segments, and?then perform a closing operation to?fill?the?holes?in the?object regions to?keep?the completeness of the object. Result To prove the validity of the proposed method, different experiment are tested on several videos, including grass, traffic section, indoor and other real scenes. All the experiments are running on the windows platform and the algorithm is implemented in C++. The adopted camera is Hikvision"s DS-2DF230IW-A with the resolution 1280×720. In order to evaluate the performanre of this method, we compare our method against?other global motion models, including affine transformation model and local linear model. The experimental results can be summarized as follows. When the frame interval is small, the affine transformation model has produced a large error, the local model and the method in this paper can achieve better results. As the rotation angle of the camera increases, nonlinear transformation will become more obvious. Because of the occurring of edge effect for the local linear model method, its compensation result will generate isolated pixels and small segments. However, the method proposed in this paper can remove 90% of the isolated pixels and small segments, which can solve the problem of nonlinear transformation for camera rotating-scan. In addition, the proposed method can be quickly solved by camera equations and Hough transform with the processing speed 50 frame/s which can meets the real-time requirements. This method also has limitations that our method?is only suitable?for low pitch angle of the camera. The influence of pitch angle on the results of our method still require further analysis and research. Conclusion Detecting moving object on a rotating-scan camera is an intractable issue because the motion of the camera leads to the movement of the background and the deformation of the image. Image compensation is required to remove background motion and image deformation. The quality of image compensation method directly affects the final result of moving object detection. For traditional methods, nonlinear transformation is not considered thoroughly. The camera imaging mechanism under the condition of camera rotating-scan is analyzed in this paper. And then a nonlinear transformation model and the corresponding calculation method are presented. Results prove that, compared with existing methods, this method can achieve real-time performance and smaller compensation error under the condition of camera rotating-scan. Based on this method, the object detection problem in the dynamic background is converted into the object detection problem in the static background, and then the reliable detection of the moving object can be achieved by using frame difference. As?pan-tilt-zoom monitoring technology is more and more widely used for scanning and monitoring of large-scale scenes, the proposed method has practical values for object detection.