Leaf Image Annotation for Accurate Insect Infestation Counting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for quantifying plant infestation by insects on leaves, such as those used by farmers, suffer from inconsistencies due to human variability in counting and are constrained by non-ideal conditions, leading to inaccurate pest management.
Innovation Solution
A computer-implemented method using convolutional neural networks (CNNs) trained with annotated images to differentiate between insect and leaf areas, followed by density map estimation to accurately count insects on leaves, utilizing both human and computer-generated annotations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If farmers visually inspect and count insects manually, then the counting process can be performed without specialized equipment, but the results vary due to human variability and lack of standardized knowledge
Solution Approach 1:
The patent replaces the manual visual inspection system with a computer vision system using deep learning models. The system processes images of plant leaves and automatically detects, classifies, and counts insects, substituting human eyes and brain processing with computational algorithms that provide consistent, repeatable measurements across different users and conditions.
Solution Approach 2:
The patent creates digital copies of insect appearances through training images and annotations. These annotated images serve as reference copies that the deep learning model learns from, enabling the system to recognize and count insects based on standardized digital representations rather than variable human perception.
2Reliability
If computer vision techniques are used to recognize insects, then objectivity and consistency improve, but the system requires large numbers of annotated training images which are time-consuming to create
Solution Approach 1:
The patent performs preliminary annotation work by having experts annotate a core set of training images before deploying the system. This upfront preparation creates a reusable dataset that can be repeatedly used to train and validate the model, avoiding the need for continuous manual annotation during operational phases.
3Measurement precision
If deep learning models are trained with expert-annotated images, then classification accuracy improves, but the cost and time of expert annotation increases
Solution Approach 1:
The patent uses a semi-supervised approach where only a portion of the training data requires expert annotation. The system combines a smaller set of expert-annotated images with a larger set of unannotated or lightly annotated images, achieving high classification accuracy while reducing the total quantity of expert annotation resources required.
Data Source
AI summary
A computer generates a training set with annotated images (473) to train a convolutional neural network (CNN). The computer receives leaf-images showing leaves and biological objects such as insects, in a first color-coding (413-A), changes the color-coding of the pixels to a second color-coding and thereby enhances the contrast (413-C), assigns pixels in the second color-coding to binary values (413-D), differentiates areas with contiguous pixels in the first binary value into non-insect areas and insect areas by an area size criterion (413-E), identifies pixel-coordinates of the insect areas with rectangular tile-areas (413-F), and annotates the leaf-images in the first color-coding by assigning the pixel-coordinates to corresponding tile-areas. The annotated image is then used to train the CNN for quantifying plant infestation by estimating the number of biological object such as insects on the leaves of plants.