Ideas for TenGun Version 4

July 15, 2025

We're thinking about what to do next for TenGun, and are seriously considering integrating OpenCV ("Open Computer Vision"). This would allow us to do a a variety of image processing tasks such as the following:

Object Detection
Detect objects in images using pretrained models (e.g., YOLO, Haar cascades) or template matching, and associate them with 3D geometry for inspection, recognition, or automation workflows.
Grayscale Image Conversion
Convert color images to grayscale to simplify filtering, segmentation, and scalar field generation based on pixel intensity.
Defect Detection
Identify cracks, holes, deformation, or surface flaws in images or texture maps using image analysis — and localize them on the 3D model for review or reporting.
Image Registration
Align two or more images of the same object taken under different conditions or from different sensors (e.g., RGB + thermal), enabling precise overlay and combined analysis.
Pose Detection
Estimate the position and orientation of an object or person from an image. Useful for augmented reality, scene understanding, or aligning 3D models with 2D reference photos.
Image-Guided Measurement & Feature Detection
Use edge or contour detection to find geometric features (holes, notches, welds) in reference images and apply annotations or measurements in the 3D scene.
Automated Screenshot & Annotation Export
Draw shapes, labels, or measurements on top of 2D screenshots using OpenCV's drawing tools, then export annotated images for reports or presentations.
Image-to-3D Projection Using Homography
Use homography to project 2D inspection images (e.g., top-down photographs) onto 3D surfaces, aligning based on keypoint correspondences.
2D Masking to Select 3D Regions
Apply masks (drawn or color-based) to isolate regions in an image, then map that selection onto 3D models or point clouds for segmentation, highlighting, or tagging.
Scalar Field Generation from Images
Convert grayscale or color images into scalar fields (e.g., heat maps) and apply them to meshes or point clouds to visualize data like thickness, temperature, or intensity.
Optical Flow for Deformation or Movement Analysis
Analyze video or frame sequences to detect motion or deformation, useful for studying material response or mechanical behavior over time.
Text, Barcode, and Label Recognition
Use OCR and barcode readers to extract identifying information (e.g., part numbers, QR codes) from scanned images and attach it to geometry or metadata.
Animated Image Sequence Projection
Load a series of images as time-varying scalar fields or textures (e.g., CT scan slices), useful for medical, thermal, or time-series visualization in 3D.

Some of these features may seem tangential to TenGun's core focus, but they unlock new ways to use the software—especially by linking simulation data with empirical results more effectively.

Update July 28, 2025

The following features are implemented so far:

Rotate, Flip, Crop image.

Grayscale and Thresholding. Useful for object recongition.

Shape (outline) detection.

AI-based background removal.

Coming soon:

Resize image.

Static AI-based object detection.

Live object and shape detection. Can lead to quality check feature.