In view of the problems that some intelligent cars can not quickly adjust the direction in the tracking process, and it is easy to deviate when passing through the curve and right angle turn, the intelligent car camera tracking system based on linear CCD is proposed and studied. The system is based on STM32F103 single-chip microcomputer, using the working characteristics of linear CCD, select TSL1401 linear CCD as a sensor to collect track information, by the single-chip PID algorithm to control the steering gear angle and motor speed, through the system hardware circuit and software programming, so that the car can identify the black guide line and according to the black line to achieve fast and stable tracking. Through field experiments, the tracking car can not only adjust the direction in time on the straight track, but also accurately and smoothly through the "8" track, turning angle of 90 ° curve. Compared with the traditional photoelectric sensor, the linear CCD sensor has a higher forward-looking, which can sense the path change earlier and correct the forward direction in advance. The use of DMA direct access technology can reduce the number of CPU processing tasks, and can directly transfer data to memory, thus improving the sampling efficiency of the system. The larger the amount of data transferred, the more obvious the efficiency improvement, which can ensure that the car can smoothly and accurately turn through larger curves. Using dynamic threshold algorithm to find out the intermediate value is more convenient for binarization of data. PID position control algorithm provides a basis for deviation correction, can make the system fast response and not overshoot. The system design improves the sensitivity and mobility of the car, basically meets the design requirements, and has reference significance for similar intelligent car design and practice.