Video Images Fire Detection Device Based On SOPC

Images fire detection technology based on video have become the new field of fire warning, an embedded video images fire detection device is introduced in this paper. It is suitable for detecting fire happen in the large space for example in the wild and warehouse etc. The device build a Nios Ⅱ system based on hardware-software co-design of SOPC using the IP provided by Altera and Avalon bus standard, it integrat image collection, storage, fire detection and network control into a piece of FPGA. The system realized the fire detection based on the dynamic characteristics of the early fire image, and the detection results and image data can be sent over the network to a remote PC, it can warning and monitoring a early fire.The device use EP3C16F256C6FPGA of Altera as the main controller, MT9M032CMOS image sensor of Micron as image collection device, W5300as network transmission chip, H57V2562GTR SDRAM as image cache that can storage running program. FPGA parallel processing and data processing advantages of Nios Ⅱ are played fully, image sensor driver and control logic of network chip are achieved using Verilog HDL, configuration of net chip, storage of image data, algorithm of fire detection and data transmission are completed by the Nios Ⅱ processor. Image data is cached to SDRAM using DMA to improve storage speed.Device analyzed and quantified the pixel gray change rule of the same position in adjacent video frames cause by the dynamic characteristics based on the change dynamic characteristics over time of fire smoke in the early fire, the change include the invariability of starting point, Overlap between frames and the overall diffusion three characteristics. first, image is preprocessed, original image is reduced three quarter by the sampling method and filter the noise using median filtering; Then suspected fire area is extracted using the improved background difference method and interframe add, and threshold segmentation and normalized are processed, and binary image is processed using morphology method in order to filter out the noise and empty, thus, computation load and storage of system are reduced; Finally, more comprehensive frame are weighted to judge whether there is a fire after the quantitative dynamic characteristics of fire smoke is extracted using the method of statistical pixel gray.Test shows the device work stablely, can achieve the early fire detection, there is better success rate, higher integration, lower cost, simple PCB wiring, and signal integrity is good.