Embedded IEEE Project Real-time Traffic Light Recognition Based on Smart Phone Platforms

ABSTRACT:

Movement light acknowledgment is of incredible hugeness for driver help or self-ruling driving. In this paper, an activity light acknowledgment framework in light of cell phone stages is proposed. Initial, an ellipsoid geometry edge demonstrate in Hue Saturation Lightness shading space is worked to separate intriguing shading locales. These districts are additionally screened with a postprocessing advance to get competitor locales that fulfill both shading and spine conditions. Second, another part work is proposed to viably consolidate two heterogeneous highlights, histograms of arranged inclinations and neighborhood paired example, which is utilized to portray the applicant areas of movement light.

Apart extraordinary learning machine (K-ELM) is intended to approve these competitor locales and at the same time perceive the stage and sort of moving lights. Besides, a spatial-fleeting examination structure in view of a limited state machine is acquainted with improve the unwavering quality of the acknowledgment of the stage and kind of activity light. At last, a model of the proposed framework is actualized on a Samsung Note 3 cell phone. To accomplish a constant computational execution of the proposed K-ELM, a CPU-GPU combination based approach is received to quicken the execution. The exploratory outcomes on various street situations demonstrate that the proposed framework can perceive activity lights precisely and quickly.

HARDWARE REQUIREMENT