Internship Presentation
KTO KARATAY UNIVERSITY
Faculty of Engineering and Natural Sciences
Department of Mechatronics Engineering
2nd Mandatory Internship
19 November 2024
Muhammed DİNÇ
Doken Technology and Machine Inc.
Content
2/17
About Company
Sensor Control Card
PCBA Comparison by Computer Vision
Node-RED
MQTT
Sipeed M1W Dock and Computer Vision
Assesment of Internship
About Company
Doken Technology and Machine Inc. (Est. 2022)
Co founders: Ozgur Ali Avcil & Mehmet Duman
Company Activities:
3/17
Figure 1. Main HQ
Seed Sensor Card
PCB Design
1809
Soldering PCBS
Plastic Injection
4/17
Figure 2. (a) PCB Boards, (b) Semi Automated Soldering Arm, (c) Plastic Injection Machine
Seed Sensor Card
Test Whole System
Single Card After Injection
5/17
Figure 3. (a) Seed Sensor Card After Molding Process, (b) Whole System Without Sensor Connections
PCBA Comparison by CV
Method | Comparison Way | Advantage | Disadvantage |
Histogram | Compares color distribution in images. | Simple and fast, works well in same object with different colors. | Doesn't capture spatial information or detailed patterns. |
Template Matching | Sliding a template over the image. | Effective for identical patterns in well-aligned images. | Sensitive to scale, rotation, and noise; computationally expensive. |
Feature Matching | Matches keypoint such as corners, edges. | Robust to scale, rotation, and partial occlusions. | Requires complex feature extraction and is computationally intensive. |
Structural Similarity Index Measure | Compare structural information such as luminance, contrast. | Captures structural differences effectively; perceptually meaningful. | Less effective with significant transformations (scale, rotation). |
Subtract | Pixel by pixel subtraction of images | Straightforward and fast for small differences. | Sensitive to alignment, noise, and global changes in lighting. |
Absolute Difference | Pixel by pixel absolute difference of images | Useful for highlighting specific differences in images. | Doesn’t account for structural or global variations. |
6/17
Table 1.Image Comparison Methods
7/17
Figure 4. Histogram Comparison Result
Figure 5. Template Matching Comparison Result
Figure 6. Feature Matching Comparison Result
Figure 7. SSIM Comparison Result
PCBA Comparison by CV
8/17
Figure 8. Subtract Comparison Result
Figure 9. Absolute Difference Comparison Result
PCBA Comparison by CV
Resize Image
ECG Algorithm
Absolute Difference
Merge Contours
Highlight Differences
9/17
Figure 10. More Results from Absolute Difference(a) (b)
(b)
PCBA Comparison by CV
Node-RED
Node-RED is a visual tool for connecting hardware, APIs, and services through flow-based programming, commonly used for prototyping and IoT applications. It enables users to create workflows with minimal coding.
10/17
Figure 11. Flow Diagram of Basic Survelliance Application
Message Queuing Telemetry Transport (MQTT)
MQTT is a lightweight messaging protocol that enables devices to communicate in real-time using a publish-subscribe model.
11/17
Figure 12. Schematic of MQTT Protocol Usage
Node-RED & Mqtt Basic App
12/17
Video 1. Basic Survelliance Application
Sipeed M1W Dock and Computer Vision
The Sipeed M1W board is designed for machine learning applications and IoT projects. It features the Kendryte K210 processor, which is optimized for edge computing tasks like real-time image processing and machine learning. The K210 includes a powerful Neural Network Processor (NPU), enabling fast AI tasks such as object and face recognition, and is ideal for low-latency, resource-efficient AI applications.
13/17
Figure 13. Sipeed M1W Dock Kit
Sipeed M1W Dock and Computer Vision
14/17
Figure 14. Summary of Trained Model
Sipeed M1W Dock and Computer Vision
15/17
Figure 15. Usage of Trained Model
Assesment Of Internship
16/17
Figure 16. All of Studies During the Internship
Thank You!
17/17