Master Thesis - Fachbereich Informatik

More documents

Recommendations

Info

5.3. EXPERIMENTAL RESULTS 115 gray level 155 150 145 140 135 130 125 120 115 110 105 0 50 100 150 200 250 300 350 400 450 500 t Mean image brightness Figure 5.10: Mean image intensity of a moving empty conveyor belt over time. The deviation between the brightest and the darkest region on the conveyor exceeds 40 gray levels and is originated in non uniform translucency characteristics of the belt. Example images showing this non uniformity can be found in Figure 3.4. as well as all other measurements belonging to this particular tube. It turns out that the average ground truth distance of this tube is 0.3mm which is still larger than the RMSE of the whole sequence due to the outliers. However it is shown that this tube is not measured wrongly in general. Furthermore one can see that all neighboring tubes that lie in the same region on the conveyor are also measured inaccurately. It is assumed that with a more uniform conveyor belt such deviations could be avoided. The mean over all measurements is 50.04 at 30m/min compared to 49.96 in the ground truth. This is still very accurate. The precision of the vision-based measurements is 0.15 compared to 0.09 of human measurements under ideal laboratory conditions. The corresponding Gaussian distributions are plotted in Figure 5.9(b). Finally, the RMSE increases with faster velocities, and the total error is larger compared to black tubes. The lowest error was measured at 20m/min (approximately the current production velocity) with 0.11. This error is still only slightly larger than the human variance. One can conclude that the results of the black tubes are very accurate both for slow and fast conveyor velocities. The RMSE falls even below the standard deviation of human measurements. The accuracy of transparent tubes decreases with faster velocities, but is still in a range that allows for measurements with the given tolerance specifications. Best results have been achieved at a velocity of 20m/min. As it turns out, all tubes meeting the tolerances in the real world (based on manual ground truth data) have been also measured reliably to be within the tolerances by the system, i.e. ΩFN =0. Thus,notubewould have been blown out wrongly at any velocity.
116 CHAPTER 5. RESULTS AND EVALUATION Diameter Ωtotal ΩPTM σtube GT Dmin GT Dmax GT D RMSE 6mm (B) 1 4.8 0.05 -0.40 0.29 -0.13 0.18 8mm (B) 1 4.6 0.05 -0.19 0.19 0.0 0.07 12mm (B) 1 4.6 0.07 -0.44 0.31 -0.11 0.19 6mm (T) 0.92 2.8 0.18 -1.15 0.87 0.01 0.20 8mm (T) 1 3.9 0.15 -0.16 0.66 0.15 0.20 12mm (T) 0.98 3.12 0.24 -0.69 0.67 0.07 0.20 Table 5.6: Measuring results of 50mm length tubes with different diameter at a velocity of 30m/min. The first two rows show to black tubes (B) and the last two rows transparent (T) ones. Figure 5.11: The thin 6mm tubes are likely to be bent. The distance between the defined measuring points in the image does not represent the length of the straight tube correctly. 5.3.4. Tube Diameter Beside tubes of 8mm diameter as investigated in the velocity experiments, there are also 6 and 12mm diameter tubes to be considered in the DERAY-SPLICEMELT series. Therefore, the test data in this scenario includes transparent and black tubes of 50mm length with these diameters. The velocity is constant at 30m/min. Again more than 100 tubes are measured for each combination of color and diameter. The summarized evaluation results can be found in Table 5.6. Black Tubes As for 8mm tubes, 100% of the black tubes both for 6 and 12mm diameter are measured by the system indicated by a score of Ωtotal = 1. The number of per tube measurements is also approximately equal with 4.8 for 6mm diameter tubes and 4.6 for 12mm tubes. The per tube standard deviation σtube is slightly larger for 12mm with 0.07 compared to 0.05 at 6 and 8mm tubes. One significant difference to 8mm tubes are the larger extrema in the ground truth distance GT Dmin and GT Dmax and the definite shift in the average ground truth distance GT D. Values of −0.13 for 6mm and −0.11 for 12mm indicate the vision-based lengths are mostly shorter than the manual measurements. This has basically two different origins: Tubes with a diameter of 6mm are bent much stronger than tubes of larger diameters as can be seen for example in Figure 5.11. In this case,bothmanualaswellasvision-basedmeasurementsaredifficult. Thelengthofatube intheimageisdefinedasthedistancebetweentheleftandrightendofthetubeatthe most outer points of the corresponding edges. If the tube is bent, however, the distance between the measuring points is obviously smaller than the real length. This can be seen in the ground truth distance as well as in the resulting RMSE which is significantly larger with 0.18 compared to 8mm tubes. Figure 5.12(a) visualizes the results of a sequence of 21 black tubes at 30m/min. The bent tube in Figure 5.11 corresponds to the 10th tube in this plot (located between measurement number 45 and 50) and is measured significantly
Page 1 and 2:
Fachbereich Informatik Department o
Page 4:
Acknowledgments First of all, I wou
Page 8 and 9:
Contents Acknowledgments iii Abstra
Page 10 and 11:
Contents ix 5.3.7. TubeLength .....
Page 12 and 13:
List of Tables 1.1. Rangeoftubetype
Page 14 and 15:
List of Figures 2.1. AccuracyandPre
Page 16 and 17:
1. Introduction Heat shrinkable tub
Page 18 and 19:
1.2. PROBLEM STATEMENT 3 hazards, r
Page 20 and 21:
1.4. RELATED WORK 5 Length [mm] Tol
Page 22 and 23:
1.5. THESIS OUTLINE 7 The vision pa
Page 24 and 25:
2. Technical Background 2.1. Visual
Page 26 and 27:
2.1. VISUAL MEASUREMENTS 11 (a) Fig
Page 28 and 29:
2.1. VISUAL MEASUREMENTS 13 Figure
Page 30 and 31:
2.1. VISUAL MEASUREMENTS 15 � xd
Page 32 and 33:
2.2. ILLUMINATION 17 tubes alternat
Page 34 and 35:
2.2. ILLUMINATION 19 effect of comp
Page 36 and 37:
2.3. EDGE DETECTION 21 i 2 i 1 0 (a
Page 38 and 39:
2.3. EDGE DETECTION 23 N × 1 kerne
Page 40 and 41:
2.3. EDGE DETECTION 25 These operat
Page 42 and 43:
2.3. EDGE DETECTION 27 (a) 0 ◦ (b
Page 44 and 45:
2.4. TEMPLATE MATCHING 29 accuracy
Page 46 and 47:
3. Hardware Configuration This chap
Page 48 and 49:
3.2. CAMERA SETUP 33 3.2. Camera se
Page 50 and 51:
3.2. CAMERA SETUP 35 eachcolorchann
Page 52 and 53:
3.2. CAMERA SETUP 37 Figure 3.4: Co
Page 54 and 55:
3.2. CAMERA SETUP 39 adjusting scre
Page 56 and 57:
3.2. CAMERA SETUP 41 further distan
Page 58 and 59:
3.3. ILLUMINATION 43 f V 12mm 129mm
Page 60 and 61:
3.3. ILLUMINATION 45 (a) (b) (c) Fi
Page 62 and 63:
3.3. ILLUMINATION 47 (a) (b) Figure
Page 64 and 65:
3.4. BLOW OUT MECHANISM 49 A light
Page 66 and 67:
4. Length Measurement Approach Whil
Page 68 and 69:
4.2. MODEL KNOWLEDGE AND ASSUMPTION
Page 70 and 71:
Page 72 and 73:
Page 74 and 75:
4.3. CAMERA CALIBRATION 59 4.2.7. B
Page 76 and 77:
4.3. CAMERA CALIBRATION 61 Figure 4
Page 78 and 79:
4.3. CAMERA CALIBRATION 63 is compu
Page 80 and 81: 4.4. TUBE LOCALIZATION 65 (a) (b) F
Page 82 and 83: 4.4. TUBE LOCALIZATION 67 � 1 Psm
Page 84 and 85: 4.4. TUBE LOCALIZATION 69 Global Th
Page 86 and 87: 4.4. TUBE LOCALIZATION 71 but also
Page 88 and 89: 4.4. TUBE LOCALIZATION 73 Global Re
Page 90 and 91: 4.5. MEASURING POINT DETECTION 75 f
Page 92 and 93: 4.5. MEASURING POINT DETECTION 77 T
Page 94 and 95: 4.5. MEASURING POINT DETECTION 79 -
Page 96 and 97: 4.5. MEASURING POINT DETECTION 81 0
Page 98 and 99: 4.5. MEASURING POINT DETECTION 83 o
Page 100 and 101: 4.5. MEASURING POINT DETECTION 85 (
Page 102 and 103: 4.5. MEASURING POINT DETECTION 87 w
Page 104 and 105: 4.6. MEASURING 89 side of the tube,
Page 106 and 107: 4.6. MEASURING 91 fcor(x) =−(c1x
Page 108 and 109: 4.7. TEACH-IN 93 inthemeasuringplan
Page 110 and 111: 4.7. TEACH-IN 95 The pair of a pixe
Page 112 and 113: 5. Results and Evaluation 5.1. Expe
Page 114 and 115: 5.1. EXPERIMENTAL DESIGN 99 between
Page 116 and 117: 5.1. EXPERIMENTAL DESIGN 101 where
Page 118 and 119: 5.1. EXPERIMENTAL DESIGN 103 Ground
Page 120 and 121: 5.2. TEST SCENARIOS 105 enough, the
Page 122 and 123: 5.3. EXPERIMENTAL RESULTS 107 The t
Page 124 and 125: 5.3. EXPERIMENTAL RESULTS 109 Detec
Page 126 and 127: 5.3. EXPERIMENTAL RESULTS 111 Lengt
Page 128 and 129: 5.3. EXPERIMENTAL RESULTS 113 GTD [
Page 134 and 135: 5.3. EXPERIMENTAL RESULTS 119 (a) (
Page 136 and 137: 5.3. EXPERIMENTAL RESULTS 121 (a) (
Page 140 and 141: 5.3. EXPERIMENTAL RESULTS 125 Total
Page 142 and 143: 5.3. EXPERIMENTAL RESULTS 127 Color
Page 144 and 145: 5.3. EXPERIMENTAL RESULTS 129 Task
Page 146 and 147: 5.4. DISCUSSION AND FUTURE WORK 131
Page 148 and 149: 6. Conclusion In this thesis a func
Page 150: Appendix 135
Page 153 and 154: 138 APPENDIX A. PROFILE ANALYSIS IM
Page 155 and 156: 140 APPENDIX A. PROFILE ANALYSIS IM
Page 157 and 158: 142 APPENDIX B. HARDWARE COMPONENTS
Page 159 and 160: 144 APPENDIX B. HARDWARE COMPONENTS
Page 161 and 162: 146 Bibliography [18] C. Demant, B.
Page 163 and 164: 148 Bibliography [51] K.K. Pingle.
show all

Master Thesis - Fachbereich Informatik

Create successful ePaper yourself

Delete template?

Save as template?