In this paper, an integrated driving environment recognition system is established on THASV-II (Tsinghua Advanced Safety Vehicle) test platform. The system collects environment information and vehicular motion information based on the fusion of machine vision, laser radar and other vehicular sensors. An adaptive bi-threshold method and steerable filters are adopted in the image pre-processing to enhance the lane marker and vehicle features. A multi-level fusion structure is designed to recognize the front vehicles and a high-order statistic method is proposed to recognize the side and rear vehicles. The anti-interference ability of lane marker recognition is improved by a point set optimization method and the target vehicle is determined through a multi-feature fusion algorithm. Field tests demonstrate that the integrated system can meet the demands of environment recognition for various lateral and longitudinal active safety systems, achieving good adaptability to environmental variation and good anti-interference ability under heavy traffic situation.