Main Content

ATS SmartVision™ Software in Medical Components Success Story

A major medical component manufacturer needed to guide a robot to pick a randomly oriented needle from a conveyor.

different sized curved needles laid out



Needles are presented in batches on a conveyor belt. A single needle had to be identified that was not touching any adjacent needles. Its location, and the position of the pointed end, needed to be identified so that a robot could pick up the needle at a specific location along its length and place it in a fixture.



A camera was placed over the conveyor. The conveyor was lit from below to illuminate the needles in profile. An inspection sequence was written in ATS SmartVision software to take a picture of the conveyor and locate a needle of the required type that was not touching any adjacent needle. The two ends of the needle were then examined to determine the sharp end from the dull end. A specified point along the needle arc was then determined and communicated to the robot for pickup.



This vision system used a calibration method to generate positional information in the world coordinate frame of the robot that was being used to pick and place the needles, simplifying the robot programming.

A blob tool was initially used to locate needles that were separated from any others. This ensured that there was no overlap which could cause a problem when being picked up by the robot. Information on the best ellipse that fit the various blobs was generated by the blob tool to determine the theoretical centre point. Edge information was then extracted from the image to refine the position of the needle and determine which end was sharp and which end was dull. The pick point was specified at a particular number of degrees of arc from the needle tip. This point was determined and the area around it was checked to ensure it was clear for the robot grippers to pick up the needle at that point.

The needles also come in “left hand” and “right hand” varieties. To improve the system cycle time, the vision system alternately located left and right hand needles.

The information on the best needle position was computed and reported directly to the robot controller over TCP/IP within approximately 150 milliseconds.


This system was successfully deployed at a customer site in France. They have since ordered an additional five duplicate systems.