TV Tracking System Based On DSP

TV Tracking System was presented in 1960s, the development of modernmicroelectric technology and computer technology makes the image tracing technologythat based on image process and Intelligent recognition developed rapidly. Diverse TVtracing devices appeared in high accuracy missile guidance, artillery control, ship-borneweapon system, target range survey, star-follower as well as the industrial automationdomain, but main application was still in high accuracy missile guidance domain. Usingtelevision image guidance is possible to provide the clear picture target, and hasadvantages in genuine and fake goal recognition, can effectively removes thedisturbance, has high reliability ,and good guidance accuracy, also because of its passivework mode, it does not launch the electromagnetic wave,and works in obviously waveband, will not be influenced by electronmagetic interference.This paper mainly introduces the research and the design of a television imagetracking system based on the DSP, and then designs and debugs the hardware andsoftware according to the main functions and the aim that the system must achieve. Thesystem needs to carry on the real-time processing of two channel input images, capturesand effectively tracks the target, and chooses one channel image to display on theterminal. Accurately computes the target position deviation in the tracking process, andoutputs it through the asynchronous serial port. At the same time, saves the deviationdata in EEPROM, and transmits them to PC with the RS232 serial port at the necessarytime, and realizes the visualization of the data.The system hardware is mainly consists of DSP, FPGA and MCU, and the mainICs are connected by I2C bus, its main function is accomplishing communication, datatransmission and hardware driving. The system uses TMS320DM642 DSP of TexasInstruments as the signal processor. DM642 is a highest-performance fixed-pointed DSP,based on the second-generation high-performance, advanced VelociTITM very-long-instruction-word (VLIW) architecture (VelociTI.2TM) developed by Texas Instruments,With the performance of up to 5760 million instructions per second (MIPS) at a clockrate of 720MHz, guarantees the system can realize the high turnover rate and largecapacity image data operation. System uses the BT656 image format for input/output, inthe three video port of DM642, VP0 and VP1 are configured as input video port,connected with Decoder SAA7115, DSP drives it through the I2C bus, and SAA7115provides the clock, HSYNC, VSYNC and FSYNC signal. SAA7115 convert the inputTV signal to digital signal, and then sends it to video port's FIFO, EDMA willautomatic stores it into SDRAM, so the core CPU of DM642 can access the SDRAM toprocess the image data.VP2 is configured as output video port, it does not directlyconnected with Encoder, but gets the image data from the output buffer and outputs it toFPGA, after OSD operation, the image data is converted to TV signal by SAA7105, andthen outputed to the display terminal. The DM642 External Memory Interface expandstwo pieces of MT48LC4M32B2 (4M*32 bit SDRAM) and a AM29LV33C (8M bitFLASH). SDRAM is connected to the 64bit data bus of EMIF, can read/write 64bit dataat every cpu cycle, be used to store program, data and image. FLASH is connected tothe low 8 bit of the EMIF data bus, be used to store bootloader and user program.System chooses the Cygnal C8051F020 MCU as control device, it uses the CIP-51micro controller shell, completely compatible with the MCS-51TM instructionset.C8051F020 connects with DM642 though the I2C bus and general purpose I/O,receives the I2C data, and outputs the data useing the two entire duplexes UART directly.At the same time, accepts the external control instruction, controls DM642 to carry onthe corresponding operation.At the software aspect, paper analysis the gate tracking, correlation tracking, multi-modes tracking, difference tracking and memory extrapolation tracking algorithm, thenchooses edge tracking and multi-modes tracking algorithm that mixed withauto-adapted gate track, correlation track and memory extrapolation track algorithm forthe two channel image input. In various edge detectors, selects the 3× 3 Sobel edgedetector and Prewitt edge detector to detect a identical image goal. Through thecomparison, we find that the Prewitt edge detector is better than Sobel edge detector,under the premise of that they both satisfy the system's real-time request, choose thePrewitt edge detector.After the system accomplished, it has carried on outdoor tracking experiment, itproved that system software and hardware work stably, the track result is okay, almostlyachieves system requirement.