Detection Algorithm Research Of Fatigue Driving Under The Infrared Video Image

Nowadays, with the rapid increase of vehicle, which has given rise to trafficaccidents occur frequently, and brought increasingly serious harm to human society.And, in a large number of traffic accidents, fatigued driving is the key factor of trafficaccidents. Hence, improving detection accuracy of fatigued driving is of importantapplication value to reduce traffic accident.In this paper, fatigued driving recognition algorithm under the infrared videoimages, which has caught out applied research, the main research work is as follows:1. The feature extraction and face detection. For the disadvantage of traditionalAdaBoost cascade algorithm such as time-consuming and poor ability ofgeneralization and so on. this paper proposes a method of preferential selection Haarfeatures?it has stronger classification ability?, to a great extent, it improves thetraining speed. At the same time, in this paper, the improved algorithm of?particleswarm optimization, PSO? combine with algorithm of AdaBoost-cascade, the paperhas built up a adaptive and generalized detection algorithm, and it can effectivelyimprove the detection accuracy and the performance of the generalized error, etc.2. Face tracking. For convergent speed is slow, texture information of gray imagesor infrared images is less, and when being tracked target happen fast moving,Mean-Shift algorithm has poor tracking shortcoming, so the paper make thecorresponding improvement to algorithm. Then, we use the SSD?Sum of Difference?algorithm to carry out global search, not only to avoid the condition of tracking failureunder fast moving, but to improve the tracking speed.3. Detection and accurate location of eyes. Through the analysis of the graygradient complexity of images to coarsely extract eye area, and?this paper utilizesthe connective algorithm of the Gabor transformation, binarization processing ofartificial threshold, gray integral projection and connected domain enhancement toimprove the accuracy of extraction, then precisely locate to the eyes; Thereby, take thecenter coordinates of eyes as a starting point, using the method of region growing toextract the size of pupil, that is, through the growth along to eight direction of itsneighborhood, then to achieve the size of pupil; Finally set radius of pupil under thenormal circumstance, and utilize the standard of PERCLOS-based measurement andfrequency characteristics of nictation under fatigue state to detect the driver's fatigue.