Automated rubber seed ventral surface identification using hue, saturation, value HSV image processing and a decision rule approach

Rubber seeds should be planted and handled correctly to boost the germination rate by placing the ventral surface facing down and adhering to the soil. Traditionally, this planting technique has been performed manually by labourers. Automation is not only the key to solving labour shortage issues but can also improve the production performance. Hence, this study was conducted to identify the dorsal and ventral surface of rubber seeds using image processing techniques of hue, saturation, value colour space and a decision rule approach. Five features were extracted at the centre of the seed based on the detected edge images, namely maximum length, ratio of major and minor axis, number of pixels, maximum convolution and number of intersections. These features were used as a dataset to develop new prediction models using a decision rule and an artificial neural network (ANN). Based on the results, it was found that the decision rule model performed better with a higher value of accuracy (88.75%), sensitivity (90%) and specificity (87.50%) compared to ANN. This was most likely due to the rules prepared by applying expert knowledge when developing a decision rule model. On the other hand, the development of the prediction model was created based on the analysis of each feature. This study could benefit the rubber industry, especially for the nursery application during the planting process, where it can potentially reduce time and labour intensity while increasing production efficiency at the same time.

Actions (login required)

View Item