PurposeThe purpose of this paper is to present a novel image‐labeling CMOS sensor for modulated marker detection.Design/methodology/approachAn image scene with multiple objects, each identified by a flashing light‐emitting diode (LED), is captured by the sensor. The LED's frequency is a representation of the object ID‐tag. The sensor detects and labels the objects by identifying the signal frequencies. The processing is performed in‐pixel and, since the object detection task is simplified, power dissipation is reduced. A 64×64 pixel sensor is designed in the 0.6 μm CMOS technology.FindingsSimulation results show successful object identification. At a frame rate of 250 fps the measured power consumption is 11 mW, which is less than those of the previously reported object detection solutions. The application of the presented sensor is shown in several different robotic fields such as unmanned aerial vehicles (UAVs) vision, household robots and industrial robots. It is also explained how the sensor can be used for low‐power localization and position detection of the robot vehicles.Originality/valueThe paper shows that the sensor is a suitable solution for low‐power landmark detection and robot localization.