Real-time Detection and Classification of Porous Bone Structures Using Image Segmentation and Opening Operation Techniques

Porous bone structures with different lattices have great potential application in medical tissue engineering as they exhibit excellent mechanical properties. In this study, we utilize the optical microscope system as the optical sensor and imager to achieve real-time detection and categorization of pores bone materials based on machine learning techniques. The initial bone images are pictured using an industrial camera, and the image processes are compiled for defining the superficial shapes of the bone configuration. The image segmentation approaches contain Canny edge detection, k-means clustering, and binarization. The initial bone surface images are transformed into the gray-scale mode, and k-means clustering is utilized to normalize the gray-scale mode for enhancing binarization precision. The erosion and dilation of the opening operation are used to extract image noises and improve the pores characteristics. The profiles and the dimensions of the pores characteristics are precisely obtained by using Canny edge detection. The Gaussian blur method is performed to acquire obvious surface profiles of the pores configurations without background noise. The experimental results show that the geometric sizes of artificial pores implants can be clearly examined by this optical microscope system after metal additive manufacturing.