I'd go with big radars in orbit.
They could also double as active SETI beacons! The collateral use would also be to generate accurate ephemerides of all solar system objects, a useful goal even if no collision hazard is detected.
If you need to see them the first time in order to protect yourself, then you probably haven't got much chance of pulling it off. But the more likely scenario is that an object may make multiple passes through the inner solar system before it presents a threat, and if we can just collect accurate enough orbital data we'll be able to predict if and when it will be dangerous in plenty of time to do something about it.