Mobileye 让视觉走进量产汽车

1999 年，希伯来大学教授沙舒亚创办了 Mobileye。这家公司做的是一件当时看起来不太性感的事：用一颗摄像头加一颗专用芯片，让量产车看懂道路。没有炫酷的无人驾驶演示，有的是一个一个车厂、一批一批前装订单，用最笨的方式把计算机视觉装进汽车。

在 Mobileye 之前，自动驾驶的算法研究热闹，但离量产很远。汽车厂商要的是可靠、低成本、能通过车规认证的感知方案，而不是实验室里的论文。沙舒亚的选择是把视觉几何的研究成果「固化」进专用芯片——EyeQ 系列——再把它打包成车道保持、碰撞预警这类具体的辅助驾驶功能卖给车企。

这条路走得很慢，却走得很深。每一代 EyeQ 芯片都在有限的算力里塞进更多算法，数以千万计的汽车搭载了 Mobileye 的方案，从欧洲到全球，ADAS 逐渐从高端配置变成安全标配。当今天人们谈论特斯拉的纯视觉路线、谈论华为的智驾方案时，Mobileye 是最早验证「摄像头+专用芯片」这条路可行的人。

Mobileye 的意义还在于它定义了自动驾驶产业链的一种商业模式：算法公司不造车，而是做「感知大脑」，与整车厂分工合作。这个模式后来被无数公司借鉴——从 ADAS 到高阶智驾，从 L2 到 L4，「算法+芯片+车厂」几乎成了行业默认的协作方式。

回看 Mobileye，它没有惊天动地的发布会，却用二十年的量产积累证明了：计算机视觉真正改变世界的方式，不一定是最前沿的论文，而是让每一项安全功能可靠地装进每一辆量产车。当自动驾驶的聚光灯亮起时，Mobileye 已经在暗处把路铺了很久。

In 1999 Hebrew University professor Shashua founded Mobileye. The company did something that looked unglamorous at the time: use one camera plus a dedicated chip to make production cars read the road. No flashy self-driving demos—just automaker by automaker, order by order, hardwiring computer vision into vehicles in the most unglamorous way.

Before Mobileye, autonomy research was lively but far from production. Automakers needed reliable, low-cost perception that passed automotive-grade certification, not lab papers. Shashua's choice was to "harden" his visual-geometry research into a dedicated chip—the EyeQ series—then package it into concrete ADAS features like lane keeping and collision warning for automakers.

The road was slow but deep. Each EyeQ generation squeezed more algorithms into limited compute; tens of millions of cars shipped with Mobileye, and ADAS gradually moved from high-end options to safety standards across Europe and the world. When people today discuss Tesla's vision-only route or Huawei's smart-driving stack, Mobileye was the first to prove "camera plus dedicated chip" was viable.

Mobileye's meaning also lies in defining a business model for the autonomy supply chain: an algorithm company that doesn't build cars but supplies the "perception brain", dividing labor with automakers. Countless companies later borrowed this pattern—from ADAS to high-level smart driving, "algorithms plus chips plus automakers" became the industry's default collaboration.

Looking back, Mobileye held no earth-shaking launch events, but two decades of production volume proved a truth: the way computer vision truly changes the world isn't always the most frontier paper—it's getting every safety function reliably into every production car. When the spotlight hit autonomous driving, Mobileye had long been quietly paving the road.