2.1. Construction Workers’ Visual Attention
As the carrier of personal feelings, emotions and behavioral information, eyes can collect and process visual information, therefore enabling people to form visual attention and spatial cognition [
16]. Eye-tracking technology is used to collect gaze, fixation and pupil metrics, which reflect people’s cognitive load and visual attention [
10,
17]. Several research efforts have been devoted to exploring the relationships among visual attention, hazard detection abilities and safety behavior. In high cognitive demand work such as decision making and hazard recognition, visual attention reflects individual perceived status and mental load, which is important to improve construction safety management [
5,
18]. Area of Interest (AOI) should be defined according to visual attention experiments, and optimizing patterns or text content can be employed to improve visual attention on AOI [
19,
20]. Eye movement process in AOI includes two types: dwelling and transition. Dwelling reflects that eye gazes in an AOI, and transition is a saccade process between different AOIs, therefore, researchers will analyze visual indicators according to different eye movement patterns in AOI [
21].
Fixation indicates a state that eyes remain still for a period of time, lasting tens of milliseconds to several seconds, which indicates personal attention on objects [
22]. Time to first fixation is an interval from experiment beginning to the first fixation on AOI, people will notice AOI faster when this time is shorter [
23]. First fixation duration indicates the time lapse of first fixation on AOI. A long duration implies that visual information is complex, since it is difficult to understand meaning of visual information in a short time [
24,
25]. For example, the first fixation duration displays the attraction of workplace hazards to construction workers’ visual attention, meanwhile, first fixation duration presents their cognition level on safety hazards [
26,
27]. Fixation metrics reflect people’s cognitive load and understanding degree. Complex visual information means more cognitive load, and people needs longer fixation duration and more counts to comprehend its meaning [
17]. Specifically, existing research analyzes construction workers’ hazard awareness by measuring fixation duration and count, finally proposes suggestions to improve their abilities on detecting workplace dangers [
6]. Saccade shows rapid eye movement in different AOIs, it reflects the changes of attention centers. Visual information can only be processed roughly as the saccade time is limited; thus, it is difficult to obtain valuable information [
6,
28]. A study illuminates that saccade can define the object shape accurately, however, it is difficult to locate its color and shape [
29]. Other studies prove that binocular saccade coordination of people with reading difficulties is low. The slow saccade speed makes them unable to recognize visual scenes quickly, so it lowers their reading efficiency [
30,
31].
In addition to fixation and saccade metrics, other eye-tracking indicators mainly include pupil size and blink counts. Pupil size reflects cognitive status, the size will change with the brightness of environment to adjust the amount of light entering eyes, to be specific, the pupil will enlarge under deep emotions [
32]. A study finds that construction workers’ pupil size in non-accident group is larger than accident group when they check potential safety hazards, and concentration makes them identify workplace hazards quickly [
33]. Blink count shows visual concentration in task implementation. Its count will decrease when the task requires high centralization, however the count will increase as attention is not concentrated enough [
34]. Another study finds that fatigue increases construction workers’ blink counts and reduces their abilities to identify dangers [
35].
To sum up, time to first fixation reflects safety signs’ visual attraction to subjects, and first fixation duration shows visual load of subjects’ first fixation. Meanwhile, fixation duration and fixation counts illustrate overall cognitive load of subjects. The above indicators can fully reflect short-term and long-term cognitive load of subjects; thus, they are selected as eye movement metrics in this study.
2.2. Construction Safety Signs
Safety signs are composed of color and geometric shape, which convey specific safety information and include three types of prohibition, compulsion and warning (ISO 3864-1 Graphical symbols-Safety colors and safety signs). Vienna Convention on Road Traffic divides safety signs into danger warning, controlling and information delivery, which are represented by triangles, circles, squares or rectangles in turn. Based on Safety Signs and Guideline for the Use, China’s safety signs contain red, yellow, blue and green, which convey the message of prohibition, warning, direction and prompt correspondingly. As an approach of safety management, they play a vital role in warning dangers and reducing accidents and casualties [
36].
In various fields, safety signs play an important role in reducing unsafe behavior and improving safety performance. For instance, relevant studies find that shape, color, symbol and text of safety signs affect people’s cognition, hence they can drive safely through understanding information meaning [
37,
38,
39]. A study from mining safety signs shows that visual attention in various locations is significantly different, and this study suggests that setting safety signs reasonably can prevent coal mining accidents effectively [
40]. Other researchers focus on safety signs’ cognitive characteristics, make people guess their meaning and score them. The results present that familiar, concrete, simple, meaningful and semantic-closed safety signs are easier for people to understand meanings [
41]. Another study on industrial safety signs exhibits that cognitive features of signs are closely related to their comprehensibility, it proposes that education and training will help people understand signs’ meaning [
42].
As an important safety measure on-site, construction safety signs convey warning information to construction workers and reduce their occupational risks and injury accidents [
9,
43]. However, the number of on-site safety signs is not sufficient, and they are not posted appropriately in dangerous areas. Moreover, the deterioration of safety signs makes it worse, since their colors cannot meet the standard requirements, and makes it difficult for construction workers to identify potential safety hazards on workplace [
44,
45]. To make full use of the warning and prompt, existing studies on construction safety signs mainly concentrate on their visual or cognitive features to propose optimization measures. In terms of visual features, a study compares safety signs optimization before and after, and finds that patterns and explanatory words enhance construction workers’ comprehension [
46]. Another study compares construction safety signs’ color standard in different countries, it finds that blue color should be reduced in workplace so as to ensure that aging workers can identify safety signs effectively [
47]. Some researchers conduct an experiment on the characteristics of safety signs, response time and counting accuracy, the results show that red and graphic content help construction workers identify and understand signs’ meaning quickly [
8]. In a study of cognitive features, the researchers ask the construction workers to draw safety signs to exhibit that their education level and spatial imagery preference effect safety sign redesign, moreover, the more specific the signs are, the easier they are to be redesigned [
48]. Furthermore, a mindset model for construction workers on understanding safety signs is established. The results present that concreteness, ease of visualization, familiarity and context availability can affect their mindset, meanwhile, construction workers with high visual imagery vividness are easier to understand meaning of abstract signs [
9].
Previous studies on construction safety signs mainly analyze the relationship between visual or cognitive features and construction workers’ comprehensibility. Specifically, visual features include color, shape, content and text; cognitive features contain concreteness, familiarity and semantic closeness. The number of studies on construction safety signs has increased in recent years, however, the number of research in the construction industry is still lower than other sectors such as transportation, public environment and occupational health [
36]. Moreover, existing studies focus on visual or cognitive features separately, the effect between the above two features on construction workers’ visual attention is neglected.
2.3. Construction Safety Signs and Eye-Tracking Experiment
According to visual information processing theory, people will form self-cognition through guessing internal structure and functional composition of things and others’ behavior [
49]. Visual information will be sent to the end-stopped neurons in visual cortex, and information processing system can be formed through deconstruction and reconstruction of PB neurons, which results in meaningful behavior [
50,
51]. Studies have shown that parietal lobe has the function of combining color shape features, fusiform and inferior temporal gyri can process visual information and form cognition [
52]. Recent studies demonstrate that people can capture and acquire visual information retinal sampling and cortical magnification [
53]. Perceptive visual information influences personal attitudes, as the improvement of information quality, it will make positive effect such as enhancing decision reliability, improving attention and making positive feedback [
13,
15,
54].
Based on visual information processing theory, when people observe an object, its characteristics are mapped to cerebral cortex through the interaction between neurons. The processing of information in brain tissues makes them recognize objects and form emotions, attitudes and behavior. At present, this theory has been applied in several fields such as marketing, education and biology. For example, a study on consumers shows that products’ color, shape, size and text attributes will influence personal visual perception and behavior, and it is a top-down process [
55]. Meanwhile, researchers from education filed find that thinking-aloud protocols can improve visual processing of text information, and increase the effect of understanding and applied learning [
14]. One biological study presents that alcohol disrupts protein fiber pathway and causes defects in visual information processing [
12]. Another study finds that lack of sleep reduces personal visual information processing rate and loses selective attention, the decline of cognitive control efficiency makes them more prone to cognitive errors [
56,
57].
Eye-tracking technology is an experimental method that uses cameras and infrared illuminators to track people’s eye movements, collect metrics including gaze position, fixation time and pupil size, it can also analyze visual information acquisition process, attention level, emotional state and subsequent behavior [
58]. With the popularity of eye-tracking technology, the number of eye-tracking experiment in construction area is increasing. For instance, a study chooses workplace safety hazards as eye movement experiment materials, the results demonstrate that construction workers with over ten-year working experience and injury exposure experience can identify safety hazards accurately and comprehensively [
26]. Further research collects eye-tracking indicators of on-site construction workers and finds that workers with high situational awareness pay more attention to tripping hazards [
59]. Additionally, work experience can improve hazard identification abilities of construction safety supervisors, therefore inexperienced supervisors should be provided with safety education and training [
60].
Safety signs can help people pay attention to potential hazards and effectively reduce workplace dangers. Therefore, it is crucial for personal safety to understand signs’ meaning correctly [
61,
62]. However, existing eye-tracking studies mainly focus on traffic signs, there is a lack of studies on construction safety signs. For instance, one study shows that once the amount of guide sign information exceeds the threshold, it will increase drivers’ cognitive load [
63]. Another research based on eye movement experiment finds that with the increase of signs’ information and cognitive strangeness, and longer gazing time is adverse to safety driving [
38]. Hence, directional signs on highway should not contain much information, and single-board signs can be used to reduce drivers’ visual workload [
37]. Meanwhile, improved traffic signs enable drivers to pay more attention to relative information timely, and they can slow down and brake [
64].
Construction safety signs are one of the most vital safety protection measures on workplace, identifying and understanding safety signs will reduce unsafe behavior of construction workers effectively. Recent studies on construction safety signs have used empirical methods such as questionnaires, case studies and semi-structured interviews to collect subjective research data, however, eye-tracking technology, EEG and other experimental methods have not been fully used to gather objective data. Since construction safety signs are consisted of visual and cognitive features, construction workers will form corresponding cognition through the processing of brain visual nerve when they note the visual information conveyed by these signs. Eye-tracking technology can collect eye movement metrics and objectively analyze the relationship between safety signs’ features and construction workers’ visual attention, which provides a more scientific basis for optimizing construction safety signs [
65]. Hence, this paper will collect eye movement indicators including time to first fixation, first fixation duration, fixation duration and fixation count [
5]. Based on the research of Chen et al. [
8] and Chan and Ng [
66], this study will explore the influencing mechanism of construction safety signs’ visual and cognitive features on visual attention through eye-tracking technology. A theoretical model is constructed based on the following hypotheses:
H1. Significant differences exist between eye movement metrics of different construction safety signs’ colors.
H2. Significant differences exist between eye movement metrics of different construction safety signs’ shapes.
H3. Significant differences exist between eye movement metrics of construction safety sign with or without auxiliary words.
H4. Significant differences exist between eye movement metrics of construction safety signs with different level of familiarity.
H5. Significant differences exist between eye movement metrics of construction safety signs with different level of concreteness.
H6. Significant differences exist between eye movement metrics of construction safety signs with different level of sematic closeness.