基于图像特征的楼梯检测算法研究

发布时间：2019-03-22 17:11

【摘要】：随着科技的快速发展,在一些生产领域和科研场所中存在高风险、高强度、高难度的工作环境,随之出现了替代人类在这些环境中工作的机器人,而楼梯状道路可能是机器人完成工作中通过的路径,这就使得机器人在该路径下要具有一定的检测识别能力。基于以上背景,本文通过单目视觉传感器获取楼梯图像信息,将楼梯边缘线作为机器人可行驶的边界,辅以相关的处理算法,进一步实现楼梯边缘检测与距离信息定位,为机器人的下一步路径规划提供一定条件。首先,针对楼梯图像在不同情况下纹理特征不同的特点,对比研究了两种特征分析方法:灰度差分统计法和灰度共生矩阵法,分别对上升和下降楼梯图像中的特征进行分析,通过两种方法对比,进而达到了区分上升下降楼梯图像的目的。其次,针对传统图像边缘检测方法中出现噪声、边缘定位不精确等缺点,本文结合小波变换和形态学的优点,提出了一种基于多方向小波变换和形态学融合的边缘检测算法,并对于不同环境下楼梯图像进行了实验,对比分析了该算法在不同环境下的适应性。然后,在楼梯图像边缘检测的基础上,利用投影变换建立摄像机与目标点的几何模型,在距离信息定位时,本文按照目标点在摄像机视野范围内的位置不同,将距离信息定位算法分成三种情况,按照目标点在成像平面坐标系中所在位置选择其对应计算公式,进一步实现楼梯边缘距离信息定位。对于计算中需要的摄像机内参数,使用了摄像机标定法中的张氏标定法进行标定。通过实验结果表明,本文提出的检测算法能够有效的实现楼梯图像的检测,提出的距离定位算法简单高效,测量误差小,为机器人下一步的运行提供有利的条件。
[Abstract]:With the rapid development of science and technology, there are high-risk, high-intensity, high-difficulty working environment in some fields of production and scientific research, along with the emergence of robots that replace human beings working in these environments. The staircase road may be the path through which the robot completes its work, which makes the robot have certain detection and recognition ability under this path. Based on the above background, this paper obtains the stair image information by monocular vision sensor, takes the stair edge line as the driving boundary of the robot, and uses the related processing algorithm to further realize the stair edge detection and the location of distance information. It provides some conditions for the next step path planning of the robot. Firstly, according to the different texture features of stair images under different conditions, two feature analysis methods, gray difference statistical method and gray level co-occurrence matrix method, are compared and studied to analyze the features of ascending and descending stair images, respectively. Through the comparison of the two methods, the purpose of distinguishing the ascending and descending staircase images is achieved. Secondly, aiming at the disadvantages of traditional image edge detection methods, such as noise and imprecise edge location, a new edge detection algorithm based on multi-direction wavelet transform and morphology fusion is proposed in this paper, which combines the advantages of wavelet transform and morphology. The experiment of stair image in different environment is carried out, and the adaptability of the algorithm in different environment is compared and analyzed. Then, on the basis of edge detection of stair image, the geometric model of camera and target point is established by using projection transformation. When the distance information is located, the position of target point in the field of view of camera is different according to the position of target point in the field of view of camera. The distance information location algorithm is divided into three cases. According to the location of the target point in the imaging plane coordinate system, the corresponding calculation formula is selected to further realize the stair edge distance information location. The camera calibration method is used to calibrate the internal parameters of the camera. The experimental results show that the detection algorithm proposed in this paper can effectively realize the detection of stair images, and the proposed range location algorithm is simple and efficient, and the measurement error is small, which provides favorable conditions for the next step of operation of the robot.
【学位授予单位】：西安科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】