Counting red grapes in vineyards by detecting specular spherical reflection peaks in RGB images obtained at night with artificial illumination

• Definition of an automatic method for counting red grapes from nightlight images.
• Detection of specular reflection peaks from the spherical surface of the grapes.
• The conclusion that the method is suitable with an average counting error of −14%.
• The number of grapes counted was mostly underestimated and with low false positives.

This paper presents an automatic method for counting red grapes from high-resolution images of vineyards taken under artificial lighting at night. The proposed method is based on detecting the specular reflection peaks from the spherical surface of the grapes. These intensity peaks are detected by means of a morphological peak detector based on the definition of one central point and several radial points. The morphological condition applied is that the intensity of the central point must be higher than all the radial points. The grape counting results obtained in different occlusion conditions were compared with a manual labeling procedure. On average, the percentage of extremely occluded grapes (occlusion higher than 75%) in the clusters was 33%, whereas the average counting detection error obtained with the automatic method proposed was −14% with only 7% of false positives, confirming that this proposal can count even highly occluded grapes.