Aquatic Identification and Sorting System (A.I.S.S) CAPSTONE PRESENTATION JOHN SCHULZ AND REINER LINTAG ADVISOR: DR. IMTIAZ MAY 4 TH , 2019 Dept. of Electrical Engineering
Agenda • Problem Statement • Architecture & Design • Results • Parts • Division of Labor 2 AQUATIC IDENTIFICATION AND SORTING SYSTEM
Problem Statement • Modular fish sorting system to address: • Invasive species • Avoid premature fish harvesting. Figure 1 “Simulation Tank” 3 AQUATIC IDENTIFICATION AND SORTING SYSTEM
Hardware Design Camera GPU:GTX 1060 ATmega128a IOT Sorting Servo Battery Powered Data Cache 4 AQUATIC IDENTIFICATION AND SORTING SYSTEM
Prototype System Design Training FPGA/ Solar Panel DNN Laptop DNN GPU Power Images Camera µC Controller Li Battery Servos Powered System Prototype Components Figure 2 “High Level System Overview” 5 AQUATIC IDENTIFICATION AND SORTING SYSTEM
Sorting Architecture Figure 3 “A.I.S.S. Architecture” 6 AQUATIC IDENTIFICATION AND SORTING SYSTEM
System Prototype Figure 4 “Current Simulation” Figure 5 “Tank Design” 7 AQUATIC IDENTIFICATION AND SORTING SYSTEM
Safety 1. Safety Management System Guidelines 1. Plan to track safety 2. Safe working environment 2. Emergency Contingency Plan 1. Accident procedure 2. Notification plan Figure 6 “GFCI Protection” 8 AQUATIC IDENTIFICATION AND SORTING SYSTEM
Results 9 AQUATIC IDENTIFICATION AND SORTING SYSTEM
Database • 20 Different lure classes Total database size: • 3 Water clarity levels +700k created in house • 3 Imaging angles captured: Images per class: ~33k Hours captured: +7 Hrs. Figure 7 “Imaging Angles” 10 AQUATIC IDENTIFICATION AND SORTING SYSTEM
Figure 8 “Database Layout” 11 AQUATIC IDENTIFICATION AND SORTING SYSTEM
12 AQUATIC IDENTIFICATION AND SORTING SYSTEM
Image Intensity • Image Intensity – the number of levels of color accuracy provided for each color • False Contouring – the appearance of lines of color that exist due to a lack of image intensity. Figure 9 “8 - Bit Color” Figure 11 “3 - Bit Color” Figure 12 “4 - Bit Color” Figure 10 “2 - Bit Color” 13 AQUATIC IDENTIFICATION AND SORTING SYSTEM
Neural Network & FPGAs • Video Card: GTX 1060 6GB Inference Performance • Training Accuracy: FP16 35,000 30,000 • FPGA Accuracy: FP12 (“FIX_12_8”) Images/Second 25,000 20,000 15,000 10,000 5,000 0 GPU GPU GPU FPGA FPGA 2016 2017 2018 2018 2019 Figure 13 “Tested FPGA Accuracy” Source: Xilinx Developers Conference 14 AQUATIC IDENTIFICATION AND SORTING SYSTEM
Neural Network • AlexNet structure • Space efficient on FPGAs • Designed for low latency time Figure 14 “Classical AlexNet” Figure 15 “Embedded AlexNet” 15 AQUATIC IDENTIFICATION AND SORTING SYSTEM
Neural Network • Optimizer: Stochastic Gradient Descent (SGD) Method Figure 16 “Training Accuracy” 16 AQUATIC IDENTIFICATION AND SORTING SYSTEM
Inferencing: Known Data Figure 17 “Verification Accuracy Known Classes” 17 AQUATIC IDENTIFICATION AND SORTING SYSTEM
Inferencing: Unknown Data Class Predictions Figure 18 “F18_ST05” Figure 19 “F12_CD09” Figure 20 “F10_CD03” Actual Class Names 18 AQUATIC IDENTIFICATION AND SORTING SYSTEM
Power Distribution • Currently using grid power for the system • Implement a maneuverable power distribution system • Portable Power Station • Car Battery • Solar Panel Figure 21 “Inverter” Figure 22 “Solar Panel” 19 AQUATIC IDENTIFICATION AND SORTING SYSTEM
Power System Power Consumption • Xilinx ZYNQ XC7Z020 FPGA – 36 [W] • Jebao DCP Water Pump – 22 [W] to 23 [W] • Mingdak LED Aquarium Light – 4 [W] System Total Wattage = 62 [W] to 63 [W] Calculation: Low = FPGA + Water Pump (Low) + LED High = FPGA + Water Pump (High) + LED 20 AQUATIC IDENTIFICATION AND SORTING SYSTEM
Power Potential Off-grid Power System Hydro-Generator • Stator • Rotor Disk • Projected Watt Generation: 59kWh Figure 23 “Stator and Rotor Model” 21 AQUATIC IDENTIFICATION AND SORTING SYSTEM
Power Potential Off-grid Power System Solar Panels • BP350 • Solar Panel: 150[W] • Solar Panel Efficiency: 24% • Portable Figure 24 “Solar Panel” 22 AQUATIC IDENTIFICATION AND SORTING SYSTEM
Parts Tank components • Jebao DCP – 2500 Water Pump • Cambro – Lid • Cambro – Container Figure 25 “Pump” • PVC Pipe ¾” • Mingdak – B00X84LQ5S Top Light • Hefty – Black Garbage Bags Figure 26 “Tank Light” 23 AQUATIC IDENTIFICATION AND SORTING SYSTEM
Parts Safety • Watt’s Wire – WW-G12T003Y GFCI • AmazonBasics - MW-A1/B3-1650 Extension Cord The team is working with and in possible water spill areas. These parts were bought and inspected before Figure 27 “GFCI” making the fully assembled tank “Live”. Safety is the number one priority for this project. 24 AQUATIC IDENTIFICATION AND SORTING SYSTEM
Parts Data Acquisition AI Training • Canon 70D DSLR camera using a 35-50mm • Alienware 13, 32GB Ram, GTX1060 6GB • Nikon D3500 DSLR camera using 18-55mm Data Storage • 1TB Samsung 970 EVO AI Implementation • PYNQ Dev. Board w/ Xilinx ZYNQ FPGA 25 AQUATIC IDENTIFICATION AND SORTING SYSTEM
Division of Labor JOHN SCHULZ REINER LINTAG Shared Labor • Build simulation tank • Build simulation tank • Testing image processing methods • Testing image processing methods • Building database • Power system • AI training and implementation • Develop electrical safety guidelines • Microcontroller & Servos • Creating website • Update Bill of Material (BOM) 26 AQUATIC IDENTIFICATION AND SORTING SYSTEM
Questions? 27 AQUATIC IDENTIFICATION AND SORTING SYSTEM
Recommend
More recommend
