Key Technology Research Of Intelligent Vehicle Monitoring System

With the development of automatic driving and the increasement ofaccidents, vehicle-mounted system based on computer intelligence playsincreasingly important role. In order to decrease the accidents caused byfatigue driving and bad vision at night. In this paper, we realize two keytechnologies of intelligent vehicle-mounted monitor system: facial keylandmark points locating and the identification of night signal lights.We improve the traditional ASM from three aspects: we add amonotonously decreasing inert factor to the update rule of shapeparameters, so as to overcome the discontinuity of shape changes due tothe hard threshold of ASM, which can enhance the stability of PCAreconstruction. Due to the deficiency of1D gradient for2D corner points,we design a Gabor feature based SVM classifier to fine tune the cornerpoints. Because ASM can’t represent the whole shape set, we build asub-D-ASM model for each of the inner local part, so as to locating theinner points accurately.For the identification of night signal lights, we analyzed the nightimage distortion due to the limit of camera dynamic scope and proposedthe locating algorithm of signal lights. Then the direction of the countdowncounter is corrected with the direction correction algorithm we proposed.At last, the content of the counter is recognized based on the measure oftemplate matching. Based on these steps, the alarm information isgenerated to remind the driver. Experiments demonstrate the effectivenessof our algorithm.