Research On Cotton Fiber Differential Detection System Based On FPGA And

As an important raw material for textile industry, quality of raw cotton directly affects its future products while the reduction of which is mainly caused by foreign fiber. In the past, the foreign fiber in the raw cotton was sorted by workers with naked eyes, which was time-consuming and costly, so it is urgent to find an efficiency way for foreign fiber detection. With the development of electronics, FPGA+DSP based image processing technology is widely used in the online inspection field. Due to the high real-time property, online inspection system based on FPGA+DSP has become the research highlight of foreign fiber detection field.Considering the different characteristics of FPGA and DSP, we used them to implement different functions:FPGA was used to raw cotton image capturing and image prejudgement, and subsequently FPGA send suspected image of foreign fibers to DSP. The DSP uses algorithms to determine whether this image has foreign fibers and then send the result back to FPGA.To synchronize the clock of image data from CCD camera to system clock, we designed a clock domain conversion module. Image prejudgement uses look up table algorithm based on synchronized SRAM, square window method to determine whether a 32×128 raw cotton image is a suspected foreign fiber image which reduced the computation of DSP.Given that DSP is the kernel device of image processing, how to send image to it and how to load and run its program is very important. This paper designed a highly adaptive bootloader for DSP when DSP and FPGA were only connected together with DSP’s EMIF interface. Meanwhile, we designed high speed image data transfer channel for DSP based on dual port RAM in FPGA and synchronous interface of DSP’s EMIF.In the end, we designed image processing algorithms for DSP which include image graying, sharpening, binaryzation and binary image filtering. Because the DSP we used is poor in floating point arithmetic and division, we designed an image graying method based on binary shift. To enlarge the differences between foreign fiber and raw cotton, we discussed two image sharpening algorithms:Sobel operator, based on first order difference and Laplace operator based on second order difference. As small impurities such as cottonseed hull is not necessary to be detected but will be determined by sharpening algorithm, this paper discussed some common binaryzation methods and proposed a window filtering method based on approximate side.