Instance segmentation using mask RCNN on chilli-padi stereo images for maturity detection

Veeran, Ranjitha (2021) Instance segmentation using mask RCNN on chilli-padi stereo images for maturity detection. Project Report. Universiti Teknikal Malaysia Melaka, Melaka, Malaysia. (Submitted)

Text (Full Text) Instance segmentation using mask RCNN on chilli-padi stereo images for maturity detection.pdf - Submitted Version Download (1MB)

Abstract

Harvesting chilli-padi at the proper stage of maturity is of paramount importance for attaining desirable quality. Conventionally, the harvesting of chilli-padi is performed manually by picking with hands which are highly time-consuming and energy-draining. Within this context, this study suggests automatically detecting the location of matured chilli-padi on a stereo vision for an automatic picking robot. This study aims to detect the chilli-padi, classify the chilli-padi by its maturity and create a mask on the detected chilli-padi images. Therefore, the Mask RCNN algorithm is an analytical framework of detecting and classifying matured and not matured chilli-padi according to the chilli-padi color. This algorithm consists of several stages to perform the detection of the chilli-padi, classification of chilli padi maturity by its color and instance segmentation on detected chilli-padi. The algorithm performance was evaluated by using a confusion matrix. From the experimental results, the algorithm can detect the chilli-padi in the stereo images. The classification achieved a prediction accuracy of 82.1% and the segmentation can create a mask on the detected chilli-padi. The experimental results show the detection, classification and segmentation can effectively perform on the chilli-padi stereo images.

Actions (login required)

View Item