Development of an image-processing technique for soil tilth sensing

Collecting, handling and sieving of the soil samples for the determination of clod size distribution is time consuming, laborious and expensive. Therefore, the aim of this paper is to use computer vision as a non-contact measurement technique for the determination of clod/aggregate size distribution in the field. Digital images were acquired from three different soil tilths, namely: coarse, intermediate and fine for sandy loam soils. Geo-correction models, digital filters and image-enhancement techniques were used in order to correct the geometric and quality distortions in the images. Three digital image-processing techniques, namely: contrast detection, edge detection and aggregate finding and classification (AFC) analysis were investigated and passed through a ‘virtual sieve’ to determine clod size distribution. Image-processing results were correlated with the results of standard sieving. The contrast detection technique was found to be significantly the best at detecting and classifying the aggregates/clods for soil tilth sensing with a size detection root-mean-square error (RMSE) of 14 mm.