Deep neural network for precision multi-band infrared image segmentation

Lu, Tingting; Huyen, Alexander; Payumo, Kevin; Figueroa, Luis; Chow, Edward; Torres, Gilbert

doi:10.1117/12.2305134

Cited by 4 publications

(4 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The experimental results are compared with the clustering methods and other segmentation methods. The clustering methods include FCM, 27 FLICM, 32 PSOFCM, 37 NLFCM, 33 and ARKFCM, 34 and the other methods include MFIS, 19 TFSSI, 47 RFLSM, 48 and GFACM. 49 The experimental results are compared with the ground truths to acquire quantitative analysis of the segmentation performance.…”

Section: Experimental Results and Analysismentioning

confidence: 99%

“…Many infrared image segmentation methods have been proposed to improve the segmentation accuracy, which could be classified into six categories, such as threshold, 8,9 mean shift, 10 Markov random field (MRF), 11,12 active contour model, [13][14][15] fuzzy C-means (FCM) clustering, [16][17][18] and neural networks (NNs). 19,20 The methods based on threshold easily produce misclassification in the image with small gray differences because the spatial information is not taken into consideration. 21 The methods based on mean shift focus on local region merging.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Infrared image segmentation based on multi-information fused fuzzy clustering method for electrical equipment

Liu

et al. 2020

International Journal of Advanced Robotic Systems

View full text Add to dashboard Cite

Serious noise pollution and background interference bring great difficulties to infrared image segmentation of electronic equipment. A novel infrared image segmentation method based on multi-information fused fuzzy clustering method is proposed in this article. Firstly, saliency detection is performed on the infrared image to obtain the saliency map, which determines the initial clustering center and enhances the contrast of the original infrared image. Secondly, the weighting exponent in the objective function is adjusted adaptively. Then local and global spatial constraints are added to the objective function of the fuzzy clustering method, which can reduce the noise and background interference. Finally, the Markov constrained field is calculated according to the initial segmentation result. After that the joint field of fuzzy clustering field and the Markov random field is constructed to obtain the optimized segmentation result. The algorithm is evaluated on the infrared images of electrical equipment, and the experimental results show that the proposed method is robust to noise and complicated background. Compared with other methods, the proposed method improves the average segmentation accuracy and T measure by about 10% and 13%.

show abstract

Section: Experimental Results and Analysismentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Infrared image segmentation based on multi-information fused fuzzy clustering method for electrical equipment

Liu

et al. 2020

International Journal of Advanced Robotic Systems

View full text Add to dashboard Cite

show abstract

“…The multi-layer nature of a DNN allows for each layer to encode distinctive features. 12 Our DNN was trained on full-color NBI and WLE images. In addition, images obtained from separating the color channels into single-channel red, blue, or green or into two-channel red + green were used to train the DNN.…”

Section: Discussionmentioning

confidence: 99%

“…For the polyp detection and segmentation, a DNN model, mask region-based convolutional neural network (Mask-RCNN), was used to identify and segment or delineate objects within images. 12 This network was used due to its performance and ability to identify objects and generate precise segmentation masks. In instance segmentation, a mask is generated around each object along with a bounding box.…”

Section: Methodsmentioning

confidence: 99%