In drone-based video sensing, the captured image stream can be understood as a temporally ordered sequence of highly correlated visual frames, where consecutive frames differ only incrementally due to the drone’s smooth motion and relatively stable environment. This continuity induces substantial redundancy, making it computationally advantageous to model frame progression in a formal, automata-theoretic framework. By conceptualizing frames as symbols in a string, the video stream can be treated analogously to a sequence of characters subjected to pattern recognition techniques such as the Knuth–Morris–Pratt (KMP) algorithm. In KMP, the presence of repeating substrings enables efficient pattern matching through the construction of partial match tables that avoid redundant computations. Similarly, in video data, repeated or near-identical frames may be interpreted as recurring “symbols” within an input sequence, suggesting a structural parallel between image repetition and substring recurrence.

An automaton defined over this sequence of frames can function as a state machine capturing the evolution of visual contexts during the drone’s flight. Each state in the automaton corresponds to a distinct visual configuration or stationary context, while transitions between states are triggered by detectable deviations in the input data, such as changes in color distribution, object presence, or spatial structure. Thus, the automaton abstracts the continuous video feed into a discrete set of states and transitions, effectively summarizing the perceptual variation encountered during the observation period.

The utility of this model lies in its ability to produce a compact representation of the entire flight. Rather than retaining every frame, which largely encodes redundant information, the automaton emphasizes transition points—moments when the state sequence changes—thereby isolating salient frames corresponding to significant environmental or positional changes. This process induces a “signature” of the flight, a compressed temporal trace that preserves the structural pattern of observed changes while discarding repetitive content.

From a computational perspective, the method provides both efficiency and interpretability. It reduces temporal redundancy by formalizing similarity relations among frames and yields a mathematically grounded representation suitable for downstream tasks such as indexing, retrieval, or anomaly detection. The resulting automaton-based abstraction thus serves as a formal mechanism for encoding, analyzing, and interpreting dynamic visual data, capturing the essential structure of the drone’s perceptual experience through the lens of automata theory and pattern matching.