Portal de Periódicos da CAPES

The automatic recognition and measurement of soil pore types by image analysis and computer programs

1984; Wiley; Volume: 35; Issue: 4 Linguagem: Inglês

10.1111/j.1365-2389.1984.tb00624.x

2056-5240

A. J. Ringrose-Voase, Paul Bullock,

Soil and Unsaturated Flow

SUMMARY A system, ANOPOR, has been developed which uses the Quantimet 720 image analyser to recognize and measure the different types of pores in impregnated soil blocks. The system is concerned with three types of pores: channels, planar voids (cracks and fissures) and vughs but is not suitable for highly interconnected pore patterns. Each of the three types has a different origin and function and presents a different two‐dimensional (2‐D) shape when sectioned. A learning set consisting of these three pore classes was used to teach the system how to recognize statistically soil pores in images using measurements made by Quantimet. To describe the pore outlines in 2‐D, shape factors giving the best class separation of the learning set in the pattern space were derived. Bayes equation was used to give the probability of a pore belonging to a particular class by comparing its position in the pattern space with the learning set. Class boundaries were determined which ANOPOR uses to allocate each pore in any image to the most likely class. The system measures the proportion of the pore space attributable to each class and the perimeter and intercept density for each. The system is used to measure the pore patterns in three soil horizons.