Due to the trend towards small lot sizes and fast changes of relevant product features in production processes, the demands for flexible measurement systems with high precision and throughput are constantly growing. By using optical scanning sensor systems, e.g. comprising a laser triangulation sensor and a fast steering mirror, the demands on speed and precision can be partially satisfied, leaving the flexibility of adapting to various measurement tasks to be solved. A common use-case is the need to measure a certain feature on a sample with a higher spatial resolution than the rest of the sample. By using machine vision such features can be detected automatically, enabling an automated adaptation of the scanning system to the measurement task. This work presents the combination of an optical scanning sensor system with an agglomerative clustering algorithm for detecting features and their dimensions. Based on the identified features, offset and scan amplitudes for high resolution rescans can be derived, resulting in a flexible metrology tool. Experimental results show that several individual features on a sample can be precisely detected and that an automatically parametrized rescan can significantly increase the lateral resolution of the acquired image.