Aviation riveting assembly error and omission automatic detection method based on multi-task learning

Abstract

The invention provides an aviation riveting assembly error and omission automatic detection method based on multi-task learning. Comprising the following steps: 1) introducing an exclusive expert module into a multi-task learning framework, extracting image features by a shared expert module and exclusive experts of each task at the same time, and fusing two different image features, finally, fusion feature images jointly extracted by the exclusive expert module and the shared expert module of each task are correspondingly output to a visual detection model and a depth estimation model respectively; 2) taking the output of the multi-task learning framework as total input image information, and embedding an ASFF (adaptive spatial feature fusion) module in a connection channel between a pyramid pooling feature fusion module and a decoder in the depth estimation model; and 3) adding two full connection layers FC behind the channel attention network channel of the visual detection model. The method has better target identification precision and depth estimation performance.

Images