Source-linked AI summary
BioLeaf: a professional mobile application to measure foliar damage caused by insect herbivory
Bruno Machado, Jonatan Orue, Mauro Arruda, Cleidimar Santos, Diogo Sarath, Wesley Goncalves, Gercina Silva, Hemerson Pistori, Antonia Roel, Jose Rodrigues-Jr
TL;DR
Traditional foliar-damage methods can be expensive, rely on laboratory facilities or complex devices, and may not handle insect-damaged leaf borders well. BioLeaf uses smartphone images, Otsu segmentation, and interactive quadratic Bezier curves to estimate damage in leaves with or without border damage. Preserved-border leaves had correlation coefficients R ≥99.76 (P-value < 0.001), while reconstructed-border leaves had R ≥99.24 (P-value < 0.001), among the six groups tested.
Problem
Traditional foliar-damage methods can be expensive, rely on laboratory facilities or complex devices, and may not handle insect-damaged leaf borders well.
Source: §1 · p. 1Method
Results
Preserved-border leaves had correlation coefficients R ≥99.76 (P-value < 0.001), while reconstructed-border leaves had R ≥99.24 (P-value < 0.001), among the six groups tested.
Source: §6 · p. 8Takeaways & Limitations
The authors conclude BioLeaf's precision is sufficient for mobile phones to substitute for expensive machinery when estimating herbivory damage.
Source: §7 · p. 9With regular lenses, the authors recommend considering leaves up to 0.5 meter.
Source: §4 · p. 6
Abstract
from arXiv · showhide
Soybean is one of the ten greatest crops in the world, answering for billion-dollar businesses every year. This crop suffers from insect herbivory that costs millions from producers. Hence, constant monitoring of the crop foliar damage is necessary to guide the application of insecticides. However, current methods to measure foliar damage are expensive and dependent on laboratory facilities, in some cases, depending on complex devices. To cope with these shortcomings, we introduce an image processing methodology to measure the foliar damage in soybean leaves. We developed a non-destructive imaging method based on two techniques, Otsu segmentation and Bezier curves, to estimate the foliar loss in leaves with or without border damage. We instantiate our methodology in a mobile application named BioLeaf, which is freely distributed for smartphone users. We experimented with real-world leaves collected from a soybean crop in Brazil. Our results demonstrated that BioLeaf achieves foliar damage quantification with precision comparable to that of human specialists. With these results, our proposal might assist soybean producers, reducing the time to measure foliar damage, reducing analytical costs, and defining a commodity application that is applicable not only to soy, but also to different crops such as cotton, bean, potato, coffee, and vegetables.
1. Introduction
Foliar herbivory quantification can inform crop-production research and decisions, but traditional and computational methods have limitations in measuring insect damage. BioLeaf is proposed as a portable, non-destructive smartphone approach that reconstructs damaged leaf borders.
1. Introduction
- Foliar area relates to photosynthesis, plant growth, and grain filling, and damage quantification can inform expert and farmer decisions, including insecticide management.The cited research uses herbivory quantification in production forecasting, plant defense, artificial defoliation, plant fitness, and ecological studies.Source: §1 · p. 1
- Traditional methods can be inaccurate, laborious, imprecise for damaged areas, or costly; area-integrating devices also perform poorly on leaves with border predation.Visual estimates may be wrong, square counting is time-consuming, and dimension-based methods cannot precisely estimate damaged areas.Source: §1 · p. 1
- Earlier computational systems measured leaf area rather than insect-damaged area; Easy Leaf Area also targets canopy area and is sensitive to illumination changes.Some cited systems additionally struggle with noise or measure only healthy leaf area.Source: §1 · pp. 1–2
- BioLeaf uses smartphone images to estimate insect damage and reconstruct lost leaf contours, enabling in-situ, non-destructive repeated measurements.The application uses quadratic Bezier curves with three user-defined control points to fit a smooth contour to damaged leaf edges.Source: §1 · p. 2
2. BioLeaf - Foliar Analysis: A Novel Approach to Estimate Leaf Area Loss
BioLeaf processes leaf images to estimate foliar loss relative to total leaf area. Its workflow combines segmentation, noise removal, interactive border reconstruction, and damage quantification.
2. BioLeaf - Foliar Analysis: A Novel Approach to Estimate Leaf Area Loss
- The four-stage workflow thresholds each image, removes noise, reconstructs damaged borders with quadratic Bezier curves, and quantifies insect herbivory.Foliar loss is calculated relative to total leaf area.Source: §2 · p. 3
- Global Otsu thresholding in CIE La*b* reduces the impact of unbalanced light reflection, and users can fine-tune unsatisfactory segmentation.The image is converted from RGB to CIE La*b* before thresholding.Source: §2 · p. 3
- Connected components remove noise by grouping adjacent pixels in binary images using 8-connectivity and filtering groups by size.The method is intended to handle small leaf pieces and sand grains remaining after capture.Source: §2 · p. 3
- For damaged borders, an expert places screen-based control points to fit one or more quadratic Bezier curves to the leaf edge.The application then counts pixels representing insect predation relative to total leaf area.Source: §2 · p. 3
3. Materials and Methods
BioLeaf’s methods combine soybean leaf experiments with image segmentation, noise removal, border reconstruction, and damage quantification. The application was tested on leaves with internal and border damage.
3.1. Experimental Design
- Soybean leaves came from natural and transgenic plants and represented caterpillar attacks, artificial defoliation, or uninjured controls; six groups yielded 36 images.The caterpillar-attack group was exposed to Spodoptera frugiperda; images were captured at 30 cm against a portable white background.Source: §3.1 · p. 3
3.2. Image segmentation by Otsu’s Method
- Otsu segmentation selects a gray-level threshold to separate two pixel classes by maximizing inter-class variance, producing a binary mask with the leaf black on white.Source: §3.2 · pp. 3–4
3.3. Bezier-based leaf reconstruction
- For damaged borders, users set Bezier control points to interpolate the missing leaf contour; the quadratic curve was effective with lower computational cost than the cubic version.The study uses three control points, with the curve endpoints placed on the leaf border.Source: §3.3 · p. 5
3.4. Application on Mobile Devices
- BioLeaf is an Android mobile application for quantifying leaf damage, evaluated across six groups of leaves and 36 images.Source: §3.4 · p. 5
- Overall correlations were R ≥99.76 for preserved borders and R ≥99.24 for reconstructed borders (both P < 0.001).For internal damage, concordance coefficients exceeded R ≥99.5; for border damage, correlation coefficients were at least R ≥98.0.Source: §3.4 · p. 5
- Internal quantification took less than 1 second per image, compared with 20 to 30 minutes per leaf for square-counting.Source: §3.4 · p. 5
4. Experimental Results
The experiments report BioLeaf’s measurement times and demonstrate non-destructive and narrow-leaf measurements. Regular lenses are recommended for leaves up to 0.5 m.
4. Experimental Results
- 10–25 seconds were needed to process images when damaged borders required reconstruction.The processing time tends to decrease as users gain experience with the application.Source: §4 · p. 6
- BioLeaf demonstrated non-destructive measurements under different illumination conditions, including undamaged leaves estimated at 0% defoliation.The examples included three plant species from the Brazilian Cerrado forestry.Source: §4 · p. 6
- For narrow-leaf plants, the application was tested on four species; with regular lenses, experts are recommended to consider leaves up to 0.5 m.Source: §4 · p. 6
5. Comparison with LI-COR 3100
BioLeaf was compared with the LI-COR 3100 across artificial and natural herbivory. The reported standard-deviation differences were small across the tested damage conditions, and the authors conclude that BioLeaf can substitute for expensive heavy equipment.
5. Comparison with LI-COR 3100
- The tests included regular artificial damage measured after approximately 25% and then 50% biomass removal, as well as irregular cuts and holes.Natural insect damage was also evaluated against the LI-COR 3100.Source: §5 · pp. 6, 8
- ±0.01% for 25% removal and ±0.02% for 50% removal were the standard-deviation differences between LI-COR 3100 and BioLeaf.The authors report linear correlation without significant divergence at any leaf size.Source: §5 · p. 6
- ±0.37% was the standard-deviation difference for artificial irregular damage, while natural herbivory showed a difference in the range of ±0.4%.Source: §5 · p. 8
- The authors conclude that BioLeaf can substitute for expensive heavy equipment used to evaluate crop herbivory.Source: §5 · p. 8
6. Discussion of the Results
BioLeaf quantified soybean foliar damage with high correlations to manual measurements for both preserved and reconstructed borders. The authors report non-destructive, on-site use, noise handling, and potential application to other crops.
6. Discussion of the Results
- R ≥99.76 with preserved borders and R ≥99.24 with reconstructed borders; both correlations had P-value < 0.001 across six groups.The comparison is with manual quantification.Source: §6 · p. 8
- BioLeaf quantified attacked areas whether damage was internal-only or affected leaf borders, while handling acquisition noise such as sand grains and leaf fragments.It does not require leaf removal, allowing repeated measurements of the same leaf.Source: §6 · p. 8
- The application ran smoothly on several mainstream phones and supports on-site image collection for further analyses.The authors report tests on Samsung S4, S6 Edge, and Sony Z2 Xperia.Source: §6 · p. 8
- The authors identify cotton, bean, potato, coffee, and vegetables as potential crops for measuring damaged areas with BioLeaf.The discussion also mentions monitoring attacks by Helicoverpa armigera.Source: §6 · p. 8
7. Conclusions
The conclusions present BioLeaf as a portable application for estimating leaf herbivory and report precision sufficient to substitute mobile phones for expensive machinery. Testing focused on soybean, with preliminary evidence of adequacy for other crops.
7. Conclusions
- The authors conclude BioLeaf's precision is sufficient for mobile phones to substitute for expensive machinery in estimating herbivory damage.The application is described as semi-automatic, interactive, multilanguage, and portable.Source: §7 · p. 9
- The methodology is instantiated as a reliable tool for in-situ leaf-damage measurement without removing leaves.The app was tested with soybean leaves.Source: §7 · p. 10
- Preliminary experiments demonstrated BioLeaf's adequacy for use with different crops, although the reported testing used soybean leaves.Source: §7 · p. 10
LI-COR 3100 (%)
The figures show regular artificial damage and irregular or natural herbivory, including correlation plots described as demonstrating high accuracy.
LI-COR 3100 (%)
- A correlation plot is described as demonstrating high accuracy for regular artificial damage at 25% and 50% damaged area.Source: Figure 9 · p. 14
- Three leaf samples illustrate artificial herbivory with irregular damage.Source: Figure 10 · p. 15
- A correlation plot is described as demonstrating high accuracy for irregular artificial damage and natural insect herbivory.Source: Figure 11 · p. 16