Dirt detection on camera module using stripe-wise background modeling

Liao, Bin; Mo, Wei

doi:10.1109/compcomm.2016.7924796

Cited by 3 publications

(2 citation statements)

References 3 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These artificial detection techniques suffer from a number of issues, including a high labor need and inconsistent sorting quality, and are substantially influenced by manufactured experiences and subjective factors. Machine vision-based detection techniques can greatly reduce the shortcomings of artificial detection.In order to find flaws, machine vision-based defect detection algorithms must first place targets in complex images, then isolate targets from the background [12][13][14][15].Liao et al [6] modeled the stripe backdrop of the image to obtain the error image. Kuo et al [7] employed the Retinex algorithm to adjust the image's gray, the traditional detection approach to identify the faults.…”

Section: Research Statusmentioning

confidence: 99%

“…Machine vision-based detection methods have been widely used in industrial fields, which can overcome, to a great extent, the defects of manual detection. Therefore, all kinds of solutions have been proposed by many scholars, including image processing methods [2][3][4][5] and machine learning methods [6][7][8][9][10][11]. However, the above-mentioned defect detection literatures have different research objects, among which the literature on image sensor is very rare.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Fast defect detection of image sensor based on lightweight network

Zhang

Yue

et al. 2023

Preprint

View full text Add to dashboard Cite

Image sensor is the main device in camera production and its quality influences the imaging quality directly. Existing detection methods can not accurately detect the defects of image sensors. To address these problems, a two-branch network based on transfer learning is proposed to solve the problem of image sensor defect detection. Firstly, model parameters and detection speed were improved by replacing the backbone feature extraction network of YOLOV with a lightweight network. Secondly, anchor box optimization and focal loss were used to solve the problem of accurate positioning and unbalanced sample classification. Finally, with the cosine annealing scheduler and fixed learning rate, the network's convergence rate and overall performance are improved. The proposed lightweight network algorithm achieves a better effect than the original algorithm. The proposed algorithm has higher detection speed and smaller parameter size than other detection methods under the same size.

show abstract