Application of Acoustic Emission and Machine Learning to Detect Codling Moth Infested Apples

Abstract

Abstract. Incidence of codling moth (CM) ( L.) infestation in apples has been a major concern in North America for decades. CM larvae bore deep into the fruit, making it unmarketable. An effective noninvasive method to detect larvae-infested apples is necessary to ensure that apples are CM-free in post-harvest processing. In this study, a novel approach using an acoustic emission (AE) system and subsequent machine learning methods was applied to classify larvae-infested apples from intact apples. ‘GoldRush’ apples were infested with CM neonates and stored at the same conditions as intact apples. The AE system was used to collect the data emitted by 80 larvae-infested and intact apples in total. Eleven AE features that changed with signaling time were obtained with the AE system. For each feature, the area under the curve along the signaling time was calculated and used as an independent input variable for the machine learning algorithms, which included linear discriminant analysis (LDA) and ensemble method adaptive boosting. With signaling times ranging from 0.5 to 120 s, classification rates for infested versus intact apples ranged from 91% to 100% for the training set and from 83% to 100% for the test set. The quick signal collection and high classification accuracy obtained in this study show the potential of AE for detecting and classifying CM-infested apples. Keywords: Acoustic emission, Apple, Codling moth, Machine learning, Pest infestation.

Non-Destructive Technologies for Detecting Insect Infestation in Fruits and Vegetables under Postharvest Conditions: A Critical Review

“…Insect pests usually bore deep into vegetable/fruit where other techniques may not be able to detect the infestation. Acoustic detection of insect activities is based on distinct sounds made by the larvae displacement when they are feeding or biochemical reactions in the pest-infested food that creates low-intensity ultrasonic sounds [17,95]. For example, crawling and feeding of two insects Callosobruchus chinensis and Callosobruchus maculatus in chickpea (Cicer arietinum) and mung bean or green gram (Vigna radiata) were monitored using a condenser-type microphone probe, with a frequency range of 20-16 KHz, placed inside an acoustic-proof bin [96].…”

“…However, most of the studies that applied an AE technique are associated with food quality attributes by mechanically destroying the food. In a recent study, Li et al [17] reported that AE detected codling moth activities in infested apples and they obtained a very high classification rate (83%) utilizing 0.5 s of acoustic signal collection. Also, an attempt was made by Ekramirad et al [105] to authenticate the specific source and signature of acoustic emission in codling moth-infested apples through correlating visually observed larvae activities, such as chewing and locomotion, with patterns in the synchronized signals from the contact Lead zirconate titanate (PZT) sensors.…”

“…The original AE system for detecting fruit fly (Drosophila) larvae activity was depicted by Litzkow et al [109], which is shown in Figure 7. Nowadays, an AE system is usually composed of acoustic sensors, preamplifier, an input-output (I/O) board, and signal preprocessing software [17]. The acoustic sensor (diaphragm) serves as the device for collecting sound signals (Figure 7), and it is placed in direct close contact with the surface of the sample, thus the acoustic signal is propagated from the sample to the sensor [110].…”

Integrated management of tortricid pests of tree fruit

“…Effective culling of fruits on packing lines is influenced by both fruit characteristics, that is colour, size of insect injury, the co-occurrence of other visual defects and operational factors, such as the numbers of fruits sorted per worker-second (Knight and Moffitt 1991). Progress in developing automated defect inspection processes based on machine vision and learning is progressing rapidly (Sofu et al 2016;Gupta et al 2018) and the possibility of detecting internal pests in apples in packinghouses has been explored but does not seem practical due to the handling times required (Li et al 2018).…”

Prediction of Radiation Frost Using Support Vector Machines Based on Micrometeorological Data

“…Generally, micrometeorological data in a frost night are multidimensional and nonlinear, which increase the difficulty in prediction. In recent years, it has been applied in solving the issues of classification, function approximation and prediction related to precision agriculture, agricultural water management and bio-meteorology [14][15][16][17][18][19]. Its successfully application in precision agriculture makes frost prediction reliable.…”