Using LiDAR and advanced algorithms to decode plant architecture for better crop yields
Imagine trying to measure the exact angle of every branch on a soybean plant across thousands of acres—a task so painstaking it would challenge the most patient plant scientist. For decades, understanding plant architecture has been more art than science, relying on crude measurements and visual estimates. Yet, soybean productivity depends significantly on how these plants grow—their branch angles determine how efficiently they capture sunlight and respond to planting density.
This agricultural challenge has now met its match through cutting-edge technology. A research team has developed SCAG, an innovative algorithm that leverages LiDAR scanning to precisely map soybean plant structure. This breakthrough in 3D phenotyping achieves what was previously impossible: accurate branch detection and angle calculation at scale, potentially revolutionizing how we develop better crop varieties 1 4 .
SCAG stands for "Stratified, Clustered, and Growing-based algorithm"—a sophisticated approach to digitizing plant architecture. At its core, SCAG transforms how we measure and understand the intricate structures of plants through three strategic processing steps:
The algorithm systematically slices the plant point cloud into horizontal layers, much like a CT scan for plants, allowing precise identification of branching points 1 .
Within each layer, SCAG groups points into meaningful clusters that correspond to individual branches or stems, effectively separating the plant into its structural components 3 .
This methodological trifecta enables researchers to move beyond simplistic measurements like plant height and canopy spread to capture complex architectural traits that directly influence plant performance and productivity.
SCAG's remarkable capabilities stem from its use of Light Detection and Ranging (LiDAR) technology, which creates detailed 3D point clouds of plants by measuring the reflection of laser pulses. Unlike photographs that capture only surface appearance, LiDAR records precise spatial relationships, mapping every stem and branch in three dimensions 1 .
The algorithm addresses three significant challenges that have long plagued automated plant phenotyping:
Thin branches that comprise few data points
Incomplete data due to occlusion
Irrelevant points that don't belong to the plant structure
By employing its stratified, clustered, and growing approach, SCAG effectively filters out noise while preserving the delicate branching structures that are crucial for understanding plant architecture 1 .
To validate their algorithm, the research team conducted an extensive evaluation across 152 diverse soybean varieties, ensuring the method would work across different genetic backgrounds and growth habits 1 4 . The experiment followed a meticulous process:
Researchers used LiDAR scanners to capture detailed 3D point clouds of mature soybean plants, ensuring comprehensive coverage from multiple angles.
Through systematic testing, the team identified ideal settings for the algorithm's three key parameters: slice height (H), growth point number (N), and slice depth (D) 8 .
SCAG processed each point cloud, identifying branches and calculating their angles relative to the main stem.
Algorithm results were compared against manual measurements taken by trained plant scientists to quantify accuracy.
This rigorous approach allowed the researchers to precisely quantify SCAG's performance while identifying its optimal operating conditions.
The experimental findings demonstrated SCAG's clear superiority over existing approaches. When evaluated against manual measurements, SCAG achieved an impressive 0.77 F-score in branch detection, significantly outperforming support vector machine (0.53 F-score) and density-based methods (0.55 F-score) 1 4 .
The algorithm's branch angle calculations showed strong correlation (r=0.84) with manual measurements, confirming its reliability for extracting biologically meaningful traits 1 . This level of accuracy is particularly remarkable given the complexity of soybean architecture and the difficulty of the task.
| Method | Branch Detection (F-score) | Angle Calculation (r-value) |
|---|---|---|
| SCAG | 0.77 | 0.84 |
| Support Vector Machine | 0.53 | Not reported |
| Density-based Method | 0.55 | Not reported |
| Crop | Angle Calculation Accuracy (r-value) |
|---|---|
| Soybean | 0.84 |
| Maize | 0.95 |
| Tomato | 0.94 |
Further testing revealed SCAG's versatility when applied to other crops. Using the Pheno4D dataset, researchers found the algorithm achieved near-perfect correlation with manual measurements for both maize (r=0.95) and tomato (r=0.94) 4 6 , suggesting broad applicability across species with different growth habits.
Parameter sensitivity analysis provided crucial insights for practical application. The research team discovered that SCAG's performance was highly robust to variations in growth point number (N) and slice depth (D), showing minimal sensitivity to these parameters. The algorithm demonstrated only moderate sensitivity to slice height (H), with optimal results achieved when H was set to approximately twice the branch diameter 8 .
Perhaps most importantly, SCAG enabled the discovery of novel plant traits with significant implications for soybean breeding. When applied to 405 soybean varieties across two consecutive growing seasons, the algorithm helped identify previously unrecognized architectural features that correlate with density tolerance 1 4 .
Two new ratios emerged as particularly promising:
The ratio of average branch angle to plant height
The ratio of average branch angle to stem length
These traits demonstrated higher heritability and repeatability than the traditional canopy width-to-height ratio (CHR), making them more reliable targets for breeding programs. Soybean lines with favorable AHR and ALR values showed enhanced ability to thrive at higher planting densities—a crucial advantage for maximizing yield per acre.
| Trait | Formula | Potential Breeding Value |
|---|---|---|
| Angle to Height Ratio (AHR) | Average branch angle/Plant height | Higher heritability for density tolerance |
| Angle to Stem Length Ratio (ALR) | Average branch angle/Stem length | Better repeatability across seasons |
| Canopy Width to Height Ratio (CHR) | Canopy width/Plant height | Traditional benchmark for comparison |
SCAG represents more than just a technical achievement—it offers tangible benefits for farmers and plant breeders alike. By enabling rapid, accurate architectural assessment, the algorithm significantly accelerates the breeding cycle, allowing scientists to identify promising lines with ideal growth habits more efficiently.
Lead researcher Shichao Jin emphasizes that the team has made their dataset, scripts, and software publicly available, ensuring that this powerful tool can benefit the entire plant science community 4 6 . This open-source approach encourages widespread adoption and continuous improvement of the method.
Looking ahead, SCAG's applications extend far beyond soybean improvement. The algorithm has potential uses in:
For desirable growth forms
For assessing tree architecture
Monitoring vegetation structure
By estimating biomass accumulation
While challenges remain—particularly in handling extremely complex plant structures or very sparse point clouds—SCAG undeniably transforms our approach to plant architecture. As we face the twin challenges of climate change and growing global food demand, such sophisticated phenotyping technologies will play an increasingly vital role in developing the crop varieties of tomorrow.
In the timeless interplay between human ingenuity and agricultural necessity, SCAG represents a powerful new ally—one that helps us see, understand, and ultimately improve the plants that sustain our world.