Author ORCID Identifier
Year of Publication
Doctor of Philosophy (PhD)
Biosystems and Agricultural Engineering
Dr. Akinbode A. Adedeji
Apple is the number one on the list of the most consumed fruits in the United States. The increasing market demand for high quality apples and the need for fast, and effective quality evaluation techniques have prompted research into the development of nondestructive evaluation methods. Codling moth (CM), Cydia pomonella L. (Lepidoptera: Tortricidae), is the most devastating pest of apples. Therefore, this dissertation is focused on the development of nondestructive methods for the detection and classification of CM-infested apples. The objective one in this study was aimed to identify and characterize the source of detectable vibro-acoustic signals coming from CM-infested apples. A novel approach was developed to correlate the larval activities to low-frequency vibro-acoustic signals, by capturing the larval activities using a digital camera while simultaneously registering the signal patterns observed in the contact piezoelectric sensors on apple surface. While the larva crawling was characterized by the low amplitude and higher frequency (around 4 Hz) signals, the chewing signals had greater amplitude and lower frequency (around 1 Hz). In objective two and three, vibro-acoustic and acoustic impulse methods were developed to classify CM-infested and healthy apples. In the first approach, the identified vibro-acoustic patterns from the infested apples were used for the classification of the CM-infested and healthy signal data. The classification accuracy was as high as 95.94% for 5 s signaling time. For the acoustic impulse method, a knocking test was performed to measure the vibration/acoustic response of the infested apple fruit to a pre-defined impulse in comparison to that of a healthy sample. The classification rate obtained was 99% for a short signaling time of 60-80 ms. In objective four, shortwave near infrared hyperspectral imaging (SWNIR HSI) in the wavelength range of 900-1700 nm was applied to detect CM infestation at the pixel level for the three apple cultivars reaching an accuracy of up to 97.4%. In objective five, the physicochemical characteristics of apples were predicted using HSI method. The results showed the correlation coefficients of prediction (Rp) up to 0.90, 0.93, 0.97, and 0.91 for SSC, firmness, pH and moisture content, respectively. Furthermore, the effect of long-term storage (20 weeks) at three different storage conditions (0 °C, 4 °C, and 10 °C) on CM infestation and the detectability of the infested apples was studied. At a constant storage temperature the detectability of infested samples remained the same for the first three months then improved in the fourth month followed by a decrease until the end of the storage. Finally, a sensor data fusion method was developed which showed an improvement in the classification performance compared to the individual methods. These findings indicated there is a high potential of acoustic and NIR HSI methods for detecting and classifying CM infestation in different apple cultivars.
Digital Object Identifier (DOI)
USDA-NIFA (Grant #: 2019-67021-29692)
Ekramirad, Nader, "Nondestructive Multivariate Classification of Codling Moth Infested Apples Using Machine Learning and Sensor Fusion" (2022). Theses and Dissertations--Biosystems and Agricultural Engineering. 92.