Velocimetry, Video, And Dsp-based Vehicle Identification

As an part of Intelligent Transport System, Vehicle Recognition & Velocity Measurement System based on video and DSP integrates some useful function, such as vehicle license plate recognition, vehicle velocity measurement, traffic supervision, toll collection, etc, and it is mainly used in supervision on over-speed vehicles.This essay will introduce what has been done by the author in the development of the system, including choosing video velocity measurement + character recognition algorithm based on wavelet transform + DSP as the system solution; development of vehicle accurate position calculation algorithm for video velocity measurement and filed test; development of fast character recognition algorithm based on wavelet transform for DSP platform and performance test; transplanting the fast character recognition algorithm from industrial control computer to DSP platform and performance test.In velocity measurement module, taking place of traditional radar, video velocity measurement use vehicle's 2D position information in the captured image to get its 3D position information in the real 3D world, and then calculate the vehicle's velocity according its 3D position information. Because of video velocity measurement's merits in prevention of electronic interference, being able to measure several vehicles' velocity at the same time, low cost, high flexibility, it will be development trend of velocity measurement technology.In character recognition module, I developed a fast alphabetic, digital and Chinese characters recognition algorithm based on Wavelet Transform. The algorithm transforms 2-Dimesion image data to 1-Dimension data by projecting the character binary image on vertical and horizon direction, as to a character image of W X H size, it reduced the operation time from about W+H level to about W X H level compared to the traditional algorithm and greatly reduced the time complexity of later processing, meanwhile it saved enough vertical and horizon lattice distribution information of the character. The 1-Dimension was decomposed by Harr Wavelet in depth of 2-layer,then we could construct a characteristic information from the low frequency part of the wavelet coefficients and compare it with the corresponding information of the standard characters, this led to a rapid and efficient character recognition. The total experimental character recognition rate was above 92%. Meanwhile, with some strong-points, such as relatively small memory, fast process speed, the fast algorithm is very suitable for application on DSP platform.Embedded system refers to the minimized, specified and realized by micro CPU system, which mainly consists of embedded CPU, relative support software, embedded operation system and application software system. Our system adopts TI DSP TMS320C6205 as main CPU to take advantage of the DSP's powerful digital signal processing capacity to realize real time vehicle license plate recognition and velocity measurement. On another aspect, owing to its low cost, good environment adaptability, high efficiency, work stability and mobility, embedded system based on DSP will also be the development direction of Vehicle Recognition & Velocity Measurement System.Including several challenging techniques, such as embedded DSP system application, optical character recognition, the real time Vehicle Recognition & Velocity Measurement System has great practical use and bright market expectation and our research also keeps ahead in domestic Vehicle Recognition & Velocity Measurement System research field.