Design and low-cost FPGA implementation of the fuzzy decision system

Methods of moving object detection for General Purpose Processors (GPPs) present an approach that does not allow a successful implementation in the Field Programmable Gate Array (FPGA). This is because correct pixel classification is directly related to the need for many resources in terms of memory or complex mathematical operations to model the reference image. Therefore, these implementations do not allow real-time processing. In addition, these methods implement only conventional techniques of digital imaging processing. Thus, we propose an architecture for FPGA that combines filter operations with fuzzy integral to improve the detection of moving vehicles. This proposed fuzzy decision system for detecting moving vehicles seeks to not compromise the results in terms of pixel classification, even validating the system through a reference image obtained by a simple modeling mechanism. Results are verified in terms of the resources occupied, maximum frequency of operation and similarity measure in the FPGA Spartan-6 LX100. The implementation of the proposed low-cost FPGA system processes high-definition images in real time and improves pixel classification.