A Binocular Vision-Based Method for Volume Measurement of Colloids

Traditional binocular vision matching methods struggle to correctly match left and right colloid images due to the high reflectivity and textureless characteristics of colloid material surfaces, making effective volume detection difficult. To address this issue, we propose a binocular vision-based method for detecting the volume of dot-shaped colloids. First, to overcome the challenge of image edge extraction caused by specular reflection on colloid surfaces, we introduce an improved method for detecting and removing specular reflections. This method effectively detects and eliminates specular highlights on the colloid surface while restoring the surface color information, thereby enhancing edge detection accuracy. Next, we acquire the morphological characteristics of colloids from two perspectives (top view and side view) using binocular vision. Finally, we construct a mathematical model of the colloid volume based on its morphological characteristics and use a scale conversion factor to calculate the colloid volume. Experimental results show that the error precision of this volume measurement method is 4.52%, meeting the requirements of industrial production.