An Aquatic Insect Imaging System to Automate Insect Classification

Sarpola, Matt J.; Paasch, Robert; Mortensen, Eric N.; Dietterich, Thomas G.; Lytle, David A.; Moldenke, Andrew R.; Shapiro, Linda G.

doi:10.13031/2013.25375

Cited by 19 publications

(16 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Once intermediate image representation using visual words are constructed, a classifier is trained to classify objects. While histogram of visual words has been applied successfully to many applications 4,19,20 , they rely on careful choice of feature size.…”

mentioning

confidence: 99%

Classification and Morphological Analysis of Vector Mosquitoes using Deep Convolutional Neural Networks

Park

Kim

Choi

et al. 2020

Sci Rep

View full text Add to dashboard Cite

Image-based automatic classification of vector mosquitoes has been investigated for decades for its practical applications such as early detection of potential mosquitoes-borne diseases. However, the classification accuracy of previous approaches has never been close to human experts' and often images of mosquitoes with certain postures and body parts, such as flatbed wings, are required to achieve good classification performance. Deep convolutional neural networks (DCNNs) are state-of-the-art approach to extracting visual features and classifying objects, and, hence, there exists great interest in applying DCNNs for the classification of vector mosquitoes from easy-to-acquire images. In this study, we investigated the capability of state-of-the-art deep learning models in classifying mosquito species having high inter-species similarity and intra-species variations. Since no off-the-shelf dataset was available capturing the variability of typical field-captured mosquitoes, we constructed a dataset with about 3,600 images of 8 mosquito species with various postures and deformation conditions. To further address data scarcity problems, we investigated the feasibility of transferring general features learned from generic dataset to the mosquito classification. Our result demonstrated that more than 97% classification accuracy can be achieved by fine-tuning general features if proper data augmentation techniques are applied together. Further, we analyzed how this high classification accuracy can be achieved by visualizing discriminative regions used by deep learning models. Our results showed that deep learning models exploit morphological features similar to those used by human experts. Mosquitoes cause global infectious disease burden, as vectors of numerous fatal diseases like malaria and dengue thrive by climate change and insecticide resistance. For example, in 2017, an estimated 219 million cases of malaria occurred and an estimated 435,000 deaths from malaria globally 1. For this reason, there have been significant efforts to develop a surveillance system for early detection and diagnosis of potential outbreaks of mosquito-borne diseases. Although, mosquito monitoring programs have been intensively developed worldwide, current mosquito monitoring procedures have many limitations. In particular, it takes at least a few days to detect potential pathogens of mosquito-borne diseases. Furthermore, one of the major bottlenecks in mosquito monitoring is that even classification by human experts of collected mosquitoes is a laborious and time-consuming procedure. Traditionally, trained researchers or technicians classify the species of mosquitoes by visual examination of morphological keys that provide step-by-step instructions on taxonomic characteristics of a given mosquito species 2. However, the number of taxonomists and classification experts has drastically decreased so far 3. Therefore, alternative automatic identification methods with expert-level classification accuracy are highly required in this field. Aut...

show abstract

mentioning

confidence: 99%

Classification and Morphological Analysis of Vector Mosquitoes using Deep Convolutional Neural Networks

Park

Kim

Choi

et al. 2020

Sci Rep

View full text Add to dashboard Cite

show abstract

“…In a vernacular manner, we will abusively use the term insects to denote these animals. The data acquired from these animals are 2D images: we do not consider for instance other kind of data, such as 3D images [19], sound [20,21], genomic data [22], etc. We are especially interested in methods that can process 2D images with "normal lighting and shooting" conditions, that is when images are acquired by standard cameras (even mobile or embedded devices).…”

Section: Scope and Previous Surveysmentioning

confidence: 99%

“…The decision of the view is taken by the human operator. In [19], a similar but more complex system is proposed: a mirror system is providing a split image, with two views of the insect approximately 90 • apart. Providing two simultaneous views increases the likelihood of good dorsal views for 2D classification, and it affords greater robustness for 3D reconstruction since the camera viewpoints of the two views are fixed.…”

Section: Category Levels Refsmentioning

confidence: 99%

A survey on image-based insect classification

Martineau

Conte

Raveaux

et al. 2017

Pattern Recognition

183

View full text Add to dashboard Cite

International audienceEntomology has had many applications in many biological domains (i.e insect counting as a biodiversity index). To meet a growing biological demand and to compensate a decreasing workforce amount, automated entomology has been around for decades. This challenge has been tackled by computer scientists as well as by biologists themselves. This survey investigates fourty-four studies on this topic and tries to give a global picture on what are the scientific locks and how the problem was addressed. Views are adopted on image capture, feature extraction, classification methods and the tested datasets. A general discussion is finally given on the questions that might still remain unsolved such as: the image capture conditions mandatory to good recognition performance, the definition of the problem and whether computer scientist should consider it as a problem in its own or just as an instance of a wider image recognition problem

show abstract

“…If the benthic macroinvertebrates are used in an extensive biomonitoring, this requires that the identification methods of the benthic macroinvertebrates are good and reliable. Generally, the automated taxa identification of benthic macroinvertebrates [7]- [10], [12]- [16], [18], [19], [23], [27] has received scant attention in the areas of pattern recognition and machine learning. Traditionally, the identification of benthic macroinvertebrate specimens is made by the taxonomist.…”

Section: Introductionmentioning

confidence: 99%

Half-Against-Half Structure with SVM and k-NN Classifiers in Benthic Macroinvertebrate Image Classification

Joutsijoki¹

2014

JCP

View full text Add to dashboard Cite

We investigated how Half-Against-Half Support Vector Machine (HAH-SVM) and Half-Against-Half k-Nearest Neighbour (HAH-KNN) methods succeed in the classification of the benthic macroinvertebrate images. Automated taxa identification of benthic macroinvertebrates is a slightly researched area and in this paper HAH-KNN was for the first time applied to this application area. The main problem, when Half-Against-Half structure is used, is to find the right way to divide the classes in nodes. This problem was solved by using two different approaches. Firstly, we applied the Scatter method for the class division problem. Secondly, we formed the class divisions in a Half-Against-Half structure by a random choice. We performed extensive experimental tests with four different feature sets and tested every feature set with seven different kernel functions in the case of HAH-SVM. Furthermore, HAH-KNN was tested with four measures. The tests showed that by the Scatter method and random choice formed HAH-SVMs performed the classification problem very well obtaining over 95% accuracy while with HAH-KNN above 92% accuracy was achieved. Moreover, the 7D and 15D feature sets together with the RBF kernel function are good choices for this classification task when HAH-SVM was used and 15D feature set, when HAH-KNN was used. Generally speaking, Half-Against-Half structure is a promising multiclass extension for SVM and an interesting variant for k-NN classifier.

show abstract

An Aquatic Insect Imaging System to Automate Insect Classification

Abstract: ABSTRACT. Population counts of aquatic insects

Cited by 19 publications

References 25 publications

Classification and Morphological Analysis of Vector Mosquitoes using Deep Convolutional Neural Networks

Classification and Morphological Analysis of Vector Mosquitoes using Deep Convolutional Neural Networks

A survey on image-based insect classification

Half-Against-Half Structure with SVM and k-NN Classifiers in Benthic Macroinvertebrate Image Classification

Contact Info

Product

Resources

About