A Novel Layer by Layer Progressive Recognition Algorithm for Wear Particle Sequence Images 核心 · 已核验

doi:10.57760/sciencedb.39878

Due to the significant variations in size and morphology among different types of wear particles, and the limited depth of field of microscopes, particles of varying thickness may appear defocused and blurred within a single ferrograph image. To address the challenges of omission and misidentification caused by defocused particles in single-image analysis, a progressive layer-by-layer recognition algorithm for wear particles in ferrograph sequence images is proposed. First, an instance segmentation model, termed WearIS, is developed for a single ferrograph image. This model incorporates the Convolutional Block Attention Module (CBAM), a variance-cascaded head network, and a segmentation branch fusing deep and shallow features to accurately identify clear wear particles in the images. Second, a progressive layer-by-layer recognition algorithm is designed to iteratively refine the recognition results across sequence images, which utilizes metrics such as the intersection-over-union (IoU) of overlapping particles between consecutive frames and confidence scores. This algorithm performs frame-by-frame association and correction, ultimately ensuring comprehensive identification of all wear particles within the sequence. Comparative experimental results demonstrate that the proposed algorithm achieves detection and segmentation AP50 values of 82.67% and 80.92%, respectively, and a mean IoU of 75.64% on the ferrograph sequence image test set, with an average processing time of 1.07 seconds per frame. Compared to single-image ferrograph analysis methods, the proposed approach significantly enhances wear particle recognition accuracy while effectively reducing the probability of omission or misidentification of anomalous particles.

落地页

https://doi.org/10.57760/sciencedb.39878

许可证

CC-BY-NC-ND-4.0 (判读置信:inferred)

国内可访问性

国内直连:可达 (2026-07-11 检测) 代理通道:可达 (2026-07-11 检测)
检测口径:lychee 双通道单轮探测;「直连超时」表示检测窗口内未完成,系慢或不稳定证据,不构成封锁证据。

故障工况

传感器