Estimating the size of densely packed and complex grains using valley edge detection with a one-pixel width and grain shape information

Abstract

This study proposes a novel image segmentation algorithm for estimating the average size of densely packed grains, particles, cells, powders, bubbles, and aggregates. The algorithm comprises two sub-algorithms: valley edge detection and grain size estimation. The valley edge detection sub-algorithm identifies weak edges among grains using four directions. In contrast, the grain size estimation sub-algorithm calculates the average size of the grains based on the detected edge density. The algorithm can recognize grains with circular, elliptical, or other regular shapes without explicitly delineating each grain, making it ideal for complex and densely packed grain images. The algorithm was tested using various samples, including aggregate particles, tomatoes, chicken/duck eggs, turtle eggs, soybeans, lawn seeds, rapeseeds, and other seeds, and the testing results were satisfactory. The proposed algorithm is 200-500 times faster than ordinary grain image segmentation algorithms. It is suitable for online applications that require real-time image processing of densely packed and detailed grain images.

Keywords:

• Grain images
• segmentation
• edge detection
• edge density
• average size

Additional information

Funding

This research is financially supported by the National Natural Science Foundation of China [grant no. 61170147], and scientific and technological projects of Henan Province, China [grant no. 202102210172].