Fast eyes detection in thermal images

Abstract

In recent years many methods have been proposed for eye detection. In some cases however, such as driver drowsiness detection, lighting conditions are so challenging that only the thermal imaging is a robust alternative to the visible light sensors. However, thermal images suffer from poor contrast and high noise, which arise due to the physical properties of the long waves processing. In this paper we propose an efficient method for eyes detection based on thermal image processing which can be successfully used in challenging environments. Image pre-processing with novel virtual high dynamic range procedure is proposed, which greatly enhances thermal image contrast and allows for more reliable computation of sparse image descriptors. The bag-of-visual-words approach with clustering was selected for final detections. We compare our method with the YOLOv3 deep learning model. Our method attains high accuracy and fast response in real conditions without computational complexity and requirement of a big dataset associated with the deep neural networks. For quantitative analysis a series of thermal video sequences were recorded in which eye locations were manually annotated. Created dataset was made publicly available on our website.

New Thermal Automotive Dataset for Object Detection

“…Knapik et al [4] presented eye detection in thermal images scheme using the virtual high dynamic range technique, to enhance performance of the dense grid of scale-invariant Redmon et al proposed a series of improved versions of the YOLO architecture, i.e. YOLOv2 [5] and YOLOv3 [6].…”

Application of the Machine Vision Technology and Infrared Thermography to the Detection of Hoof Diseases in Dairy Cows: A Review

“…The highest-resolution thermal image for the sources examined was 640 × 512 pixels. Another characteristic drawback of IRT images is low contrast and a significant amount of noise, with the typical feature of IRT images being blurred edges of the observed object contrary to the image in the visual range [57,63].…”

“…Resolution of Thermal Images [64] 60 × 80 (0.005 MP) [65,66] 160 × 120 (0.019 MP) [67][68][69][70][71][72][73][74][75][76][77][78][79][80] 320 × 240 (0.077 MP) [63,81,82] 320 × 256 (0.085 MP) [83,84] 336 × 256 (0.086 MP) [85,86] 384 × 288 (0.111 MP) [59][60][61][62][87][88][89][90][91][92][93][94] 640 × 480 (0.307 MP) [95,96] 640 × 512 (0.328 MP)…”

“…The number of clusters can be quite low, for example, three [112], four [113], or six [80]. Knapik and Cyganek used K-means segmentation to reduce the number of prototype symptom descriptors [63]. Another type of segmentation method is edge detection [92], with its typical examples including Sobel, Prewitt, or Canny edge detectors [66,114].…”

Yüz İmgelerinden Göz Bölgelerinin Tespitinde ESA Tabanlı Alternatif Bir Yaklaşım

“…Termal görüntülerin kontrastını arttırmak için yeni sanal yüksek dinamik aralık prosedürü ile bir ön işleme önermiştir. Önerilen yöntem hızı ve doğruluğu YOLOv3 derin sinir ağı tabanlı nesne dedektörü ile karşılaştırılarak ortaya konmuştur [12]. Gou C. vd.…”