首页 > 行业新闻

北京中科慧眼科技有限公司

2016/07/02

随着各国政府和消费者安全性意识大幅提升,单车智能提升是实现无人驾驶的必经阶段,以ADAS(高级辅助驾驶)为代表的智能驾驶产业将率先爆发。因此,我公司进行ADAS的研发工作完全符合政府对汽车驾乘安全的要求,顺应目前市场潮流,以创新性自主研发的基于双目立体视觉的ADAS系统,让广大用户都能享受科技带来的安全、智能驾车新体验。

    基于双目立体视觉的ADAS系统,将相机拍摄到的景物实时转换为距离,无须大规模数据采集帮助机器学习,可以实现前方不规则障碍物的碰撞预警,后期工作量非常小,精度可达到厘米级,使用过程无需标定。该系统使汽车从被动安全转向主动安全,提高汽车驾驶安全性,大幅减少交通事故的数量和减轻事故伤害的严重程度。

    系统具备防碰撞预警(人、车、障碍物)功能;车道偏离预警。 并具有1080P,160度视角,30fps行车记录仪功能;可连接手机APP,将行车数据、用户信息、故障数据、报警数据通过手机APP上传并保存;系统具有车载Wifi功能。

系统采用基于多尺度多路径Viterbi算法的立体视觉算法;应用自动驾驶自校准技术,针对自动驾驶时传感器安装到驾驶室内,得到的图像会受到多重因素干扰(1.光线要透过驾驶室的斜面玻璃;2. 摄像头支架的抖动;)的问题提出的解决方案。解决实时的图像自校准,能够消除上面两个不利因素对于视差图的影响。基于立体视觉的环境感知,以双目立体视觉算法解决智能驾驶中环境感知的应用问题,主要包含:(a)研究基于立体视觉的小物体检测、路沿算法;(b)研究基于立体视觉的检测路面的算法;(c)研究基于立体视觉以及GPS地图的交通路牌及信号灯的识别;(d)研究基于立体视觉的行人、车辆检测。基于CUDA的智能驾驶实时算法,由于采用双目立体视觉技术,因此计算量非常大,为了解决这个问题,对于自校准技术、立体视觉的环境感知进行CUDA优化,实现实时计算。

本系统目标市场是汽车后装市场个人客户,还可作为前装与国内各车企合作,成为车企供应商,为汽车提供主动安全预警。
 

As governments and consumers have increased security awareness, intelligent car is to achieve the necessary stage to achieve unmanned, ADAS (advanced auxiliary driving) as the representative of the intelligent driving industry will take the lead in the outbreak. Therefore, our company were ADAS R & D work in full compliance with the request of the government of automobile driving safety, conform to the current market trends, to innovation of independent research and development of the ADAS system based on binocular stereo vision, let users can enjoy the technology can bring security, intelligent driving experience.

The ADAS system based on binocular stereo vision, camera photographed the scene real-time converted to a distance, without large-scale data collection to help machine learning, can be achieved ahead of irregular obstacles Collision Warning, later workload is very small, precision can reach centimeter level ,the use of the process needs no calibration. The system makes the car from the passive safety to active safety, improve the safety of vehicle driving, significantly reduce the number of traffic accidents and reduce the severity of accident injury.

The system has the function of anti collision warning (human, vehicle and obstacle), Lanedeparture warning system. And with 1080p, 160 degree angle, 30fps tachograph; connect the phone app, the vehicle data, user information, fault data, alarm data by mobile phone app to upload and store. The system has on-board WiFi function.

The system based on multi-scale and multi path Viterbi algorithm of stereo vision algorithm; the application of automatic driving self calibration technique, for autopilot sensors mounted to the cab, the image will be affected by the interference of multiple factors and put forward solutions.(light to through the cab beveled glass;Camera support jitter). To solve real-time image self calibration, to eliminate the impact of the above two unfavorable factors on the disparity map. Based on stereo vision perception of the environment, the binocular stereo vision algorithms solve problems in the application of the intelligent driving environment perception, mainly includes: (a) the study of based on stereo vision of the small object detection, road along the algorithm; (b) of algorithm based on stereo vision detection of pavement; (C) study on the stereo vision and GPS map of traffic signs and signals recognition based on; (d) to study based on stereo vision of pedestrian and vehicle detection. Based on CUDA intelligent driving real time algorithm, due to the adoption of the binocular stereo vision technology, the calculation amount is very large, in order to solve this problem, for self calibration and stereo visual perception of the environment of CUDA optimization, realize the real-time computation.

The system target market is the individual customers market, but also can be used as a front mounted with the automobile enterprises to become suppliers of automobile