An Advanced Method of Acquiring Distorted Images Information with Height Compensation

The innovation work is to acquire distorted images information with height compensation. Computer vision allows observation of the surroundings. Generally, images collected from cameras in automated production lines are front views. But there is a certain angel between the camera and the working platform in limited space which will result in perspective distortion and make following works such as feature extraction and object recognition much more difficulty. Even more, it will reduce precision for grasping objects and production efficiency. For efficient considerations, correction of the distorted image has potential economic and social needs. Thus, we propose the combined method of Hough transformation and perspective transformation, which can convert the distorted image into a front view. In addition, we remove the effect of objects’ height to improve positioning accuracy. Then we apply it to the Delta robot to correct the distorted image and recognize objects. It is important to note that the camera can be mounted in any position. Verification of the effectiveness of method is required for a final practical test of the experiment. The experimental results have proved the accuracy of this method well. Moreover, positioning accuracy can be easily improved by an average of more than 10 percent with the method of height compensation. Introduction As a numerical control tool, robots with the characteristics of machine vision occupy a large proportion of life and industry. The transportation robot was used in the field of logistics automation initially, plays an important role in industrial manufacturing automation. Most of the transfer robots are designed and preprogrammed for Teaching-Platform. The starting and ending positions of this type of robot must be first immobilized. The motion trajectory of robot has been predetermined. The objects are not recognized practically. With the implementation of computer vision (Ibaraki and Tanizawa, 2011; Li et al., 2015; Saegusa et al., 2010), we can solve this problem. In order to accomplish more complex tasks, it is necessary to recognize the location of objects. Typical robot can select unordered objects on the conveyor before, then place on the next belt regularly. Tho and Thinh (Tho and Thinh, 2015) and Zhang et al. (Zhang et al., 2012) have proposed an algorithm for detecting circles with cameras that are perpendicular to the working platform. Liu et al. (Liu et al., 2016) acquire the information of the road centerline accurately from a distorted image. The proposed algorithm based on inverse perspective transformation. Koufogiannis et al. (Koufogiannis et al., 2011) present an image rectification framework using perspective transformation to require accurate knowledge of the internal image structure. The framework is used for the automatic rectification, metric correction, and rotation of distorted integral images. Malkov et