Study On Stair Detection Algorithm Based On Image Feature

With the rapid development of science and technology,there are many high-risk,high-intensity,difficult work environments in some production areas and scientific research places.As a result,Robots are beginning to replace humans work in these high-risk environments.The staircase may be the robot through the path to complete the work,which makes the robot to have a certain degree of independent detection and navigation capabilities to achieve the path.Based on the above background,this paper obtains the staircase image information from the monocular vision sensor,and the edge of stair line as the boundary of the robot,supplemented by the related processing algorithm to realize the stair edge detection and distance information positioning.It provides the next path planning for the robot certain condition.First of all,according to the different characteristics of the stairs in different situations,this paper uses two feature analysis methods:Gray-Scale Difference Statistical Method and Gray Coefficient Matrix Method.The characteristics of the ascending and descending staircase images are analyzed separately.We achieve the purpose of distinguishing the ascending and descending staircase image through the comparison of the two methods.Secondly,in view of the traditional image edge detection method noise,edge location is not accurate and other shortcomings.This paper proposes an algorithm of wavelet transform and morphological fusion edge detection algorithm,which based on the advantages of wavelet transform and morphology.In this paper,we make the experiments on stair image in different environments,and compared and analyzed the adaptability of the algorithm in different environments.Then,This paper completes the stairway distance information positioning on the basis of the edge of detection.In this paper,the geometry measurement model between camera and target is built according to the perspective projection.When measured,the camera field of view is divided into three regions.It is proposed to use the scanning method combined with the distance measurement model to achieve the distance calculation.For the camera calibration method for calibration.The result of experimentation shows that the algorithm proposed in the thesis can realize the detection of stair image effectively.The visual measurement presented in this paper is simple,low calculation and accurate.So,to sum up,this provides some conditions for the next step in the operation of the robot.