Image Processing

1) Payload Release Looping

            This is the procedural setup for activating the payload and navigation system. The integration of image processing block diagram from surveillance sensor to flight control system consists of the object recognition algorithm.

            The algorithm is divided into image segmentation steps that take the colour and texture information into account followed by a model-object detection step which includes the shape, scale and context information. The image is then matched with predefined target and used for data association and tracking with data acquired from GPS. 

              Lastly, an overriding command is sent to flight controller to activate the release mechanism as well as overriding the flight mode.

2) Image Segmentation

The image segmentation stage divides the original image into three different classes based on the colour and texture features. The images are collected in Red Green Blue (RGB) colour space. They are then transformed into a Hue, Saturation and Value (HSV) colour space to reduce the sensitivity towards change in light intensity.

After feature extraction the colour and texture features are grouped into one single feature vector consisting of three colour channels and thirty texture channels. Each feature vector is then assigned with a label representing its class. The aim is to segment the original image into three different classes, object, shadow and background.

The classifier converts the original colour images into images with meaningful class labels.

3) Object Model Detection

Object detection algorithms based purely on statistical information have their performance limited by the quality of data. The algorithm used to generate a target object outline is used as the prior knowledge for object detection where a simple round shape is used to approximate the edge of ground target known as edge matching.  Any potential object detections with the wrong shape and edges can be rejected. 

4) Image Matching

The image matching technique applied here is based on image segmentation with RGB values and objects model definition with shape and edge matching. This method is useful to detect predefined ground target. The higher the UAV flies, the more structure from the environment can be captured and thus, image registration is more reliable at higher altitude.

After both methods have been processed, a matching algorithm tries to identify the best match with the predefined image. The image obtained with highest percentage of matching in terms of colour, shape and edges will validate the criteria set for the activation of release mechanism.