Research Of Image Processing Technology For Structured Light Stripe Image

Three-dimensional visual inspection based on structured light has been more and more widely used because of its high precision, easy to extract the image information, real-time and other characteristics. Structured light image processing is one key task of the entire detection process in structured light visual inspection system. It is difficult to deal with structured light image processing for two reasons. One is that the subject has a wide variety of geometric shapes and textures, and the other is that there are different kinds of shooting method and lighting conditions. For above two reasons, we can neither use only one structured light image processing method for all kinds of structured light images, nor use basic image processing methods for structured light images. Based on the above requirements, this paper researched image processing technology for structured light stripe image, and to achieve it. The main work of this paper is as follows.(1) As the impact of uneven illumination in the process of image feature information extraction, this paper presents an image processing algorithm that can both eliminate the impact of uneven illumination and maintain the main characteristics of images. In order to accelerate the algorithm, we use separable filters to reduce the complexity of the algorithm. Then we propose another accelerated algorithm based on CUDA. Experiments show that the algorithm can eliminate the impact of the illumination on structured light images. The stripes can be clearly distinguished and easily divided.(2) Due to the lack of image detail in the dark parts which caused by various factors, we proposed an improved Retinex algorithm with color restoration. Experiments show that the algorithm can recover detail information in dark parts while maintaining the original color information. When applying this algorithm on structured light images can make the stripes in the shadow parts re-emerging. This facilitates follow-up treatment.(3) Highlight information and texture information not only increase the difficulty of extracting structured light stripe, but also affect the accuracy of the resulting data. This paper proposed a highlight parts detection method based on K-Means algorithm and a stripe segmentation method based on minimum skewness method for solving the problems caused by high information. We also proposed another stripe segmentation method based on noise pixel blocks detection combined with frame difference method for solving the problems caused by color information.(4) In this paper, we researched the fingerprint image enhancement methods. We proposed an improved stripe segmentation algorithm based on fingerprint image enhancement methods. Experiments show that the algorithm can effectively enhance the stripe information of structured light images.(5) In this paper, we designed and implemented an image information processing system which can segment stripes from structured light stripe images. This system includes all algorithms we proposed above, and also provides easy-to-use user-editing tools.