Quantitative characterization of plant architectural complexity is critical for understanding genotype-to-phenotype relationships and supporting modern crop improvement. Conventional phenotyping metrics, such as Leaf Area Index and Fractal Dimension, capture isolated geometric properties but fail to represent the integrated structural organization of plant architecture.

This thesis presents an image-based phenotyping framework for quantifying plant structural complexity and enabling genotype discrimination using interpretable architectural descriptors. A Structural Complexity Index (SCI) is first introduced by integrating complementary shape-based and skeleton-based features extracted from RGB images. Principal Component Analysis is employed to summarize high-dimensional morphological information into a single interpretable score, enabling scalable comparison across genotypes and growth stages.

To improve genotype discrimination, a multi-descriptor phenotyping pipeline is developed that combines geometric, skeletal, and structural traits with kernel-based non-linear representation learning and supervised classifiers. This framework captures conditional, combinatorial, and higher-order interactions among architectural features, enabling improved genotype separation beyond traditional linear metrics.

Finally, topological data analysis using Persistent Homology is incorporated to encode global structural organization, including branching connectivity and internal voids. The integration of topological descriptors with geometric features yields robust phenotypic representations that enhance genotype classification accuracy and resilience to noise and occlusion.

Overall, this work establishes a scalable and interpretable framework for topology-aware plant phenotyping and genotype-to-phenotype analysis.