ABSTRACT
Machine vision technology began to be widely implemented in fruit packhouses in the 1970s, replacing grading machines based on mechanical principles (diverging belt for size grading, counterbalance weight tipping buckets for weight grading, and the human eye for color grading). By the 1990s, image classification functions (e.g., using a neural network-based classifier) were being employed for surface defect detection on fruit moving along the pack line, distinguishing between acceptable features (e.g., a stalk) and blemishes. A major challenge in this application was speed of object movement. With pack line speeds of 1 m/s common, a given fruit passes through the camera image area at 1 mm/ms. Mechanisms to rotate the fruit during passage through the image area, and use of angled side
16.6 Image processing-Spectral, thermal, and others ........................... 358 16.6.1 Thermal or thermal and RGB imaging .................................. 359 16.6.2 Spectral imaging ....................................................................... 359 16.6.3 Canopy estimations .................................................................. 359
16.7 Comparative performances of imaging techniques ........................ 366 16.7.1 Fruit detection rate .................................................................... 366 16.7.2 Robustness ................................................................................. 366
16.8 Summary and recommendations ....................................................... 368 16.8.1 Imaging sensors ........................................................................ 368 16.8.2 Time of day ................................................................................ 368 16.8.3 Image processing ...................................................................... 369 16.8.4 Incorporation to a horticultural system ................................. 369
Acknowledgment ........................................................................................... 370 References ........................................................................................................ 370
mirrors to reveal the sides of the fruit to the camera, were innovations to improve machine vision performance.
