The invention discloses a
smoke sensing test method based on depth camera
perception and mechanical arm control, and the method comprises the steps: obtaining a
smoke rod image collected by a depth camera at the
tail end of a mechanical arm, and obtaining a
smoke sensing result based on camera calibration and hand-eye calibration in combination with a multi-channel
cascade decision model for confidence evaluation and a priority-guided dual-mode threshold
search algorithm. A grabbing calibration object on the tobacco rod is positioned; the mechanical arm is controlled to complete tobacco rod grabbing, and the mechanical arm is positioned to move to the initial position of the smoke
detector; acquiring the optimal capturing
pose estimation of the tobacco stem reaching the smoke
detector; the tobacco stems grabbed by the mechanical arm are accurately controlled to move to the smoke
detector; and after obtaining the
test response that the tobacco stem alignment test meets the specified
waiting time, the mechanical arm drives the tobacco stem to return to the initial position. According to the
system disclosed by the invention, the full-process
automation of the smoke sensing test from tool preparation, target positioning, precise alignment to test reset is realized.