مقاله انگلیسی رایگان در مورد تشخیص چهره خودکار میمون رزوس – الزویر 2018

Automated methods for monitoring behavior of laboratory animals such as rodents and zebrafish are becoming widespread (Bains et al., 2016; Nema et al., 2016; Steele et al., 2007). They allow the monitoring of behavior effects in response to experi mental manipulations such as drugs, lesions, genetic modifications and disease. Concurrently similar automated behavior systems are being developed for monitoring health and welfare in a range of species including farm and laboratory animals (Roughan et al., 2009; Rushen et al., 2012). Rhesus macaques are one of the most common non-human primate species used in biomedical research including neuroscience research but to date the use of automated systems with non-human primates has been limited. A major challenge in measuring the behavior of any grouphoused animal is to reliably identify an individual animal. With macaques this is not an issue if the animals are singly-housed but welfare concerns are driving a move towards pair- and grouphousing of non-human primates in many countries. One solution is to add a tracking device to each animal; for example these could be colored jackets (Rose et al., 2012) or collars (Ballesta et al., 2014) in combination with video monitoring or electronic devices such as RFID tags (Maddali et al., 2013). These require regular handling of the animals and it is not currently known how the use of jackets and collars affects the behavior of the animals (personal observations with rhesus macaques suggest that the use of jackets can drastically reduce social behaviors such as grooming). Another solution is to use biometric identification based on the distinguishing visual characteristics of that species (e.g. coat pattern; Kühl and Burghardt, 2016). This has the advantage of being non-invasive. Rhesus macaques, in common with many primate species, do not have obvious individually identifiable features but the macaques themselves are capable of recognizing conspecifics by their faces (Parr et al., 2000). Face recognition technology has already been used in several non-human primate species including guenons (Allen and Higham, 2015), chimpanzees (Freytag et al., 2016) and gorillas (Loos, 2012) but not rhesus macaques. Freytag et al. (2016) achieved success rates of over 90% with images of captive chimpanzees. Face recognition technology was originally developed for use with humans and is becoming commonplace in daily life. Uses include automatic passport gates at airports, tagging of faces in photos on Facebook and use of facial image to unlock smart phones. Many of the early techniques focused on either reducing the dimensionality of the facial image or on extracting a particular feature from the image and then on classifying this output. Some of these methods for face recognition include EigenFaces (based on principal component analysis) and FisherFaces (based on linear discriminant analysis; Belhumeur et al., 1997). Some of the main challenges facing any face recognition system are coping with changes in light intensity and pose. A method based on local binary patterns (Ahonen et al., 2006) has been shown to be relatively robust to changes in light intensity. Most recently deep learning techniques have been applied to face recognition with a high level of success (Freytag et al., 2016 for chimpanzees; Parkhi et al., 2015 for humans).