The stochastic microstructure of recycled carbon fibre-reinforced composites poses significant challenges for process optimization and performance prediction, due to the disparity in scales between fibre diameter and fibre agglomerates size. This study establishes a comprehensive quantitative characterization framework to evaluate the microstructural evolution of recycled carbon fibre thermoplastic tapes processed under two distinct melt-compounding regimes with contrasting processing conditions. Utilizing a deep learning-based EfficientNet-B4 model and computer vision for robust fibre segmentation, we analysed the interplay between fibre morphology, orientation, and dispersion on large-field high-resolution micrographs. A novel quantitative approach revealed microstructural notions, hard to detect otherwise, and uncovered a critical trade-off governed by shear mixing history: while a shorter processing scenario preserved fibre length (>600μm), it resulted in a planar-random orientation state driven by the tumbling of large aggregates. Conversely, the longer processing scenario induced significant fibre fragmentation but facilitated a transition towards uniaxial alignment. To identify these changes in microstructure, novel homogenization factors (Cluster Prevalence Indicator (CPI) μ and Intra-Cluster Density Indicator (IDI) κ) and a Collective Similarity Index (CSI) were introduced. These metrics demonstrated that extended mixing transitions the material from a state dominated by volumetrically large, internally dense super-clusters to a locally dispersed microstructure, without altering the invariant global spatial periodicity (d^∗≈260μm). Furthermore, a comparative evaluation of statistical descriptors highlighted that while conventional topological metrics (e.g., Voronoi statistics) proved insensitive to processing variations, the proposed cluster-based metrics and eigenvalue-based orientation tensors successfully captured the structural nuances. This framework provides essential methods for tailoring the processing-structure relationships in recycled discontinuous fibre composites. ... Read more