A VLSI DBSCAN processor composed as an array of micro agents having self-growing interconnects

A DBSCAN (Density Based Spatial Clustering of Applications with Noise) processor has been developed. The original DBSCAN algorithm has been divided into two parts: the self-growing process and the labeling process for efficient VLSI hardware implementation. An array of micro agents having self-growing interconnects has been developed to achieve full pixel-parallel processing for self-growing as well as for labeling by single clock-cycle connection. A proof-of-concept chip was designed in a 0.18-μm CMOS technology. From the simulation results, a frame rate of 12K fps is demonstrated even for the worst case processing of an image of 80 × 80 pixels at an operating frequency of 100 MHz.