Method for discriminating of the shape of overlapped apple fruit images

Abstract

Listen

Translate article

The discrimination the shape of overlapped apple images was investigated. A combination of procedures was used to form the process using basic image processing to obtain a complete fruit region within the collected apple image. The connected region in the fruit image was framed by horizontal minimum rectangle and the Euclidean distance method was used to calculate the distance of each fruit in the connected region. The grey value of distance image increases from the outside to the inside, and becomes the maximum in the centre of the fruit region. Thus, a horizontal and vertical projection curve was obtained. Using the improved local extremum method, the effective peaks were extracted from the curve smoothed by the moving average method. The shape of the overlapped apple was discriminated based on the number of peaks. To verify the effectiveness of algorithm, 30 apple images including overlapped fruits were selected from captured images for experiments. Result showed that the algorithm effectively discriminated overlapped shape of apple fruits, which was superior to the previous methods.