Phenotyping remains a bottleneck in many breeding programs, including sweet cherry. Current fruit evaluation protocols require extensive manual sorting and visual evaluation, which reduces throughput and is subject to evaluator bias and fatigue. The Washington State University Cherry Breeding Program is seeking more efficient methods of evaluating fruit quality. In 2023, the program acquired an optical fruit sorter. Our objective was to customize the sorter parameters according to breeding program needs and compare the results of the sorter with traditional methods. Our Tomra InVision 2 sorter has the same optics, software and computer hardware as a commercial sorter, but operates on a single lane. Fruit are loaded onto an infeed system which passes fruit in single file into the detection area. A combination of fruit rotation, multiple cameras and mirrors is designed to image the entire surface of individual fruit. Both visual and infrared images are captured, generally > 24 images per fruit at a rate of approximately 15 fruit per second. The sorter software identifies fruit and classifies them according to a set of tunable quality parameters or grades. Air-actuated valves then eject the fruit into one of four grade-determined exits. The sorter generates reports that include the fruit size profile as well as the percentage of fruit sorted into the various exits and/or grades. The sorter shipped with a pre-loaded map (sorting algorithm), which we modified by updating with data from representative images of various quality parameters. We then used the sorter to grade fruit from Phase 2 variety trials. We analyzed 50-fruit subsamples in the traditional manner for size and defects. All remaining fruit from each sample were analyzed via the sorter. Out of 20 samples evaluated, the average number of fruit per sample evaluated by the sorter was 154, vs. 50 for manual evaluation. Overall, the sorter detected a lower percentage of cracking, doubles (polycarpy), and pitting vs. manual evaluation, and a higher percentage of skin blemishes. Continued testing will be required to determine whether these differences are due to the effects of small sample size or bias due to the methods themselves (human evaluator vs. sorter). While the sorter required a similar number of personnel as for manual evaluation, it required less time to evaluate each sample even though more fruit were analyzed. We will expanding the use and testing of the sorter in 2024, including evaluation of postharvest quality.
