Vision-Based Object Detection & Localization

Key Engineering Contributions

1. 01
   Color Segmentation: Programmed an automated HSV thresholding algorithm to isolate target colors. Explicitly handled hue wrap-around logic mathematically to ensure robust target detection without data loss.
2. 02
   Blob Analysis & Filtering: Implemented morphological operations to clean raw binary masks, effectively removing pixel noise and filling internal gaps. Utilized connected component labeling to extract precise target centroids and pixel areas.
3. 03
   Geometric Transformation: Formulated mathematical mappings to convert 2D pixel coordinates into real-world spatial estimations. Calculated the target's angular offset based on the camera's 60° horizontal field of view and estimated relative proximity.
4. 04
   System Validation: Validated the detection algorithm against static training datasets, unseen test environments, and dynamic webcam feeds, proving the system's reliability.

Visual Documentation

Figure 1

01.png

HSV color thresholding in MATLAB

Figure 1

01.png

Pipeline of cone detection starting from the original image, binary image conversion via multiple thresholding techniques, and finally proximity estimation for the mobile robot’s decision making