A Novel Automated Inspection Approach Based on Adaptive Region-Growing Image Segmentation

以適應性區域成長影像分割為基礎的創新自動化檢測方法

林存國(Tsun-Kuo Lin)

《中國機械工程學刊》 35卷1期 (2014/02) Pp. 57-65

摘要

區域成長演算法由於可用種子像素辨識影像區域，因此常用於影像分割技術。本研究提出一創新區域成長技術用於自動化檢測。此創新技術是以類神經網路為基礎的適應性區域成長演算。此演算將檢測影像轉成灰階影像，之後依據類神經網路的不變矩影像形狀因子辨識結果，作區域影像分割。此方法能自動產生分割影像與最佳化形狀因子作為影像檢測。此方法在自動化檢測試驗效能良好，與其他影像分割方法比較，有較佳的結果。

並列摘要

The region-growing algorithm is commonly used for image segmentation because the algorithm can identify regions by selecting seed points. This study presents a novel algorithm for adaptive region growing based on neural networks, which is highly effective as a region-growing technique for automated inspection. The algorithm transforms input images into a gray-level space and then adaptively segments the images by merging regions based on artificial neural networks, which classify the image patterns according to shape descriptors of moment-based invariants. This approach can automatically produce segmented images with optimal shape descriptors for inspection. The proposed method performs well in automated inspection tests and produces superior results to existing methods of image segmentation.

並列關鍵字

adaptive image segmentation ； image inspection ； neural networks ； region growing