Download presentation
Presentation is loading. Please wait.
Published byMohamed Goodson Modified over 9 years ago
1
LU Ming YAN Rui YU Jia CHEN Ying QIU Qiangqiang
2
Introduction Parameter setting Experiment Design Data analysis and conclusion Further improvement
3
Background Our purpose Introduction
4
Electronic traffic signs, especially based on LED, seem to an overwhelming trend. Dynamic Updated Connectable Solar energy supported Background
5
Purpose Our experiment is to study the color factors (combination, brightness) influencing visual acuity on electronic panel under high-speed condition.
6
The meaning of our experiment To identify the factors that have significant influence on the short-time LED discernation. The main application of LED for short time discernation is transportation area. Model To simulate the real driving condition is the base stone to assure the validation of our experiment. Physical Meaning of Parameters
7
Parameter setting Color system Preparation
8
The relationship between vehicle velocity and reaction time. Noise simulation To simulate the disturbance and noise in driving, play touring video at the background Equipment configuration--- consistency The angle of computer LCD The screen brightness The angle of the chair for the participants Simulation Condition
9
The relationship between vehicle velocity and reaction time. Reaction time=Discerning time + Action time The time for the driver to discern the traffic signs. Display time, The time to display the picture. Discerning time the time for driver to take actions responding to the signs Answer time,the time to answer the question Action time Comparative, easy to simulate Not comparative, clicking the mouse& steering wheel Velocity Simulation
10
Velocity(km/h)406080100 Average Reaction Distance(m) 435411374326 Average Reaction Time(s) 3.01.91.30.9 Table 1 Reaction distance to traffic sign under different velocities Reaction time= Discerning time + Action time High road Discernation time Scale =100 ms Velocity (km/hr) Display time (s) Answer time (s) Value assigned1000.10.8 Velocity Simulation
11
Color Combination http://en.wikipedia.org/wiki/Munsell_color_system Hue: Each horizontal circle Munsell divided into five principal hues: Red, Yellow, Green, Blue, and Purple Value, or brightness, varies vertically along the color solid, from black (value 0) at the bottom, to white (value 10) at the top. Five color: R, B, Y, G, P Two level of Brightness: 4, 6 Five color: R, B, Y, G, P Two level of Brightness: 4, 6
12
Picture Treatment Simple to recognize Pictures The difficulty to recognize is similar Color RGB 25500 00 0 0 0 0
13
Picture Treatment Five colors Two kinds of brightness: high and low Low High
14
Picture Treatment Background Picture … Total 400 pictures Each kind of color combination have 20 pictures,10 with high brightness and 10 with low brightness 20 kinds of color combination
15
Experiment Design
16
FactorsHigh LevelLow Level Brightness64 GenderFemaleMale Content ColorFive color: red, blue, yellow, green, purple Background ColorFive color: red, blue, yellow, green, purple Almost like a 4-factor full factorial design 40 Pictures (randomly chosen from 400 pictures) per testee (include all the combination of the color and the brightness) Gender Plan to have 20 female testees and 20 male testees, that is 20 replications
17
Experiment Testees18—22 PlaceIn the dormitory EquipmentComputer Picture Display Time0.1s Screen LightlessLargest possible Angel of Screen, distance, environmental noise, temperature, light Fixed (dormitory environment) Graph characteristicsFixed Testees tirednessAlmost the same Experiment Condition Control
18
Selection Area Control Area Picture Display Area Introduction of Experiment Software
19
Selection of raw data Graphically analysis and quantification analysis Binary choice model Data Analysis
20
MaleFemale 2320 Eliminate the data sets that are not full as 40 Eliminate some data sets so that design is Orthogonal (the data sets of male is equal to the sets of female ) MaleFemale 19 Finally 1520 data to be analysis Eliminate the data sets that are obviously outliers Raw Data Selection
21
All people’s data sets are in control which mean differences between testees are acceptable There are two full score among the 38 data sets (see as the R graph) Difference Between testees
22
N: 38 (19+19) Mean: 35.5 Std: 2.565 Normal distribution fitted AD: 0.509 P-Value: 0.186>0.05 Normal distribution fitted Descriptive data analysis
23
Purple—Yellow 0.7895 Purple—Yellow 0.7895 Yellow---Blue 0.9474 Yellow---Blue 0.9474 Green---Yellow 0.9474 Green---Yellow 0.9474 RedYellowBlueGreen Purple Simple view of color combinations
24
Quantification of Factors and Model Building Background Color Content ColorBrightnessGenderAnswer (response) R1G1B1R2G2B2 6 (High) 4 (Low) RGB Red25500 Blue00255 Green02550 Yellow255 0 Purple2550
25
Main factors and interactions
26
Factorial Analysis 拟合因子 : 答案 与 R1, G1, B1, R2, G2, B2, 亮度, 性别 项 效应 系数 系数标准误 T P 常量 0.88980 0.012772 69.67 0.000 R1 -0.04934 -0.02467 0.011061 -2.23 0.026 B1 -0.05378 -0.02689 0.015247 -1.76 0.078 R2 -0.04934 -0.02467 0.011061 -2.23 0.026 亮度 -0.03289 -0.01645 0.010429 -1.58 0.115 性别 0.05044 0.02522 0.010429 2.42 0.016 R1* 性别 -0.03728 -0.01864 0.009328 -2.00 0.046 G2* 亮度 -0.04364 -0.02182 0.012339 -1.77 0.077 B2* 亮度 -0.04408 -0.02204 0.012339 -1.79 0.074 S = 0.314934 PRESS = 173.692 R-Sq = 2.75% R-Sq (预测) = 0.00% R-Sq (调整) = 0.73% 对于 答案 方差分析(已编码单位) 来源 自由度 Seq SS Adj SS Adj MS F P 主效应 8 1.785 2.131 0.26642 2.69 0.006 2 因子交互作用 23 2.392 2.392 0.10401 1.05 0.399 残差误差 1488 147.585 147.585 0.09918 失拟 48 5.585 5.585 0.11636 1.18 0.189 纯误差 1440 142.000 142.000 0.09861 合计 1519 151.763
27
拟合因子 : 答案 与 R1, B1, R2, 性别 项 效应 系数 系数标准误 T P 常量 0.89151 0.008625 103.37 0.000 R1 -0.04007 -0.02004 0.008625 -2.32 0.020 B1 -0.03430 -0.01715 0.008358 -2.05 0.040 R2 -0.03436 -0.01718 0.008512 -2.02 0.044 性别 0.03816 0.01908 0.008075 2.36 0.018 S = 0.314811 PRESS = 151.134 R-Sq = 1.07% R-Sq (预测) = 0.41% R-Sq (调整) = 0.80% 对于 答案 方差分析(已编码单位) 来源 自由度 Seq SS Adj SS Adj MS F P 主效应 4 1.617 1.617 0.40414 4.08 0.003 残差误差 1515 150.146 150.146 0.09911 失拟 11 0.903 0.903 0.08205 0.83 0.613 纯误差 1504 149.243 149.243 0.09923 合计 1519 151.763 Factorial Analysis
28
响应曲面回归 : 答案 与 R1, G1, B1, R2, G2, B2, 亮度, 性别 项 系数 系数标准误 T P 常量 0.891447 0.010432 85.450 0.000 R1 -0.021272 0.009331 -2.280 0.023 G1 -0.005373 0.012344 -0.435 0.663 B1 -0.019189 0.012344 -1.555 0.120 R2 -0.016009 0.009331 -1.716 0.086 G2 -0.000110 0.012344 -0.009 0.993 B2 0.007127 0.012344 0.577 0.564 亮度 -0.007237 0.008081 -0.896 0.371 性别 0.019079 0.008081 2.361 0.018 S = 0.315051 PRESS = 151.776 R-Sq = 1.18% R-Sq (预测) = 0.00% R-Sq (调整) = 0.65% 对于 答案 的方差分析 来源 自由度 Seq SS Adj SS Adj MS F P 回归 8 1.785 1.785 0.22314 2.25 0.022 线性 8 1.785 1.785 0.22314 2.25 0.022 残差误差 1511 149.977 149.977 0.09926 失拟 71 7.977 7.977 0.11236 1.14 0.205 纯误差 1440 142.000 142.000 0.09861 合计 1519 151.763 Response Surface Analysis
29
响应优化 目标 下限 望目 上限 权重 重要性 答案 望目 0.8 0.95 1 1 1 全局解 R1 = 0.0020920 G1 = 255.000 B1 = 0.0016753 R2 = 255 G2 = 254.998 B2 = 254.998 亮度 2 = 4.27523 性别 = 0.0158782 预测的响应 答案 = 0.95, 合意性 = 1.000000 复合合意性 = 1.000000 Response Optimization Background Color—Green Content Color—White Brightness—Low Gender—Male
30
Problem of the Model Response Variable Y is Binary {0, 1}, not continuous. Disobey the classic hypotheses listed as below: 1.Y=XB+N. XB+N is continuous. Y is discrete. The equation itself 不成立。 2. N~ normal( 0, ) so Y= normal(, ) But actually, Y is not normally distributed. 3.1520 sample, most of them (1349/1520) are y=1. Minitab regards y=0 is outliers. The ordinary modeling method is not validated. Binary Choice Model is a better fit.
31
Binary Response Model Analysis with EViews Significant factors: R1, Gender, R2 Probability (LR stat) :0.0169: Model is significant
32
Binary Response Model ANSWER = 1-@CNORM(-(1.622034779 - 0.0008928323457*R1 - 0.0001714847162*G1 - 0.000752830862*B1 - 0.0006876612784*R2 - 2.709367975e-06*G2 + 0.0003340929489*B2 - 0.03865378463*BRIGHTNESS + 0.2169980384*GENDER)) Probability
33
Binary Response Model Fitness Model is properly fitted!
34
Residual Plot P=0.5
35
Analysis Flow 1 Raw data selection to get the valid data sets 2 Graphically analysis 3 Quantification of the factors Factorial analysis Response surface analysis and optimal solution 4 Binary choice model with EViews Final model
36
Conclusion 1.The color combination of Yellow—Blue and Green—Yellow perform well in discerning while Purple—Yellow performs badly. Green background performs better. 2.The significant factors are R1, B1, R2, GENDER 3.The best combination of all factors are: Background: Green Content: White Brightness: Low Gender: Male 4.A binary response model is properly fitted with the situation and can be used to predict the probability of correct discerning.
37
Further Improvement
38
Due to technological and economic limits, there are some shortcomings in our experiment. a.Undermine the verification and validation of our experiment b.Affect the accuracy of our sample data c.Break the consistency between our sample and our goal. Some can be improved while others not. Further Improvement
39
Things can be improved The unexpected appearance of signs Expected clicks the “next” button and the next picture will be displayed. DOE Unexpected never know when the traffic sign will show up Reality Show time as random generated Further Improvement
40
Tense nerve of driver Tense nerve is caused by the sense of high speed, which is hard to simulate. It has a significant impact on the identification and decision process. Fatigue Driving always lasts hours long. The visual fatigue cannot be simulated in the short experiment. Experience of driver The familiarity of traffic signs and driving experience influence the discernation process. Environment variations The air visibility cannot be simulated. Things cannot improved Further Improvement
41
EFFECTS OF COLOR COMBINATION ON VISUAL ACUITY AND DISPLAY QUALITY WITH TFT-LCD, Journal of the Chinese Institute of Industrial Engineers, Vol. 23, No. 2, pp. 91-99 (2006) Effects of bending curvature and text/background color-combinations of e-paper on subjects’ visual performance and subjective preferences under various ambient illuminance conditions ScienceDirect, Displays 28 (2007) 161– 166 Measurement of Human Sensation for Developing Sensible Textiles Human Factors and Ergonomics in Manufacturing, Vol. 19 (2) 168 – 176 (2009)
42
THANK YOU THANK YOU
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.