In collaboration with Gonzalo Arocena and Marcos Aitcin.

CDT is a system based on digital image processing for studying the process of fruit ripening. As an application of this system a prototype for analyzing the ripening process of tomatoes was built. This system acquires images of the whole surface of the fruit, processes that sequence of images and generates a pseudocylindrical equal area projection in which the area of regions of interest can be measured. The evolution of the area of these regions of interest is used for studying the fruit ripening process.

The prototype has three main modules: Capture, processing and analysis.

In order to capture in detail the entire surface of the tomato, the system is equipped with a stepper motor capable of performing rotations of 1.8 °. An application developed in C++ using openframeworks synchronizes the rotation of the motor with the acquisition of images. Each time the motor moves one step the application captures a new image. The fruit to be analyzed is located over the motor using a special mount that transmits the torque without causing slips. Thus, when the motor completes a 360 degree turn a sequence of 200 images is acquired. The driver of the stepper motor was implemented using an arduino board.

The processing module was developed in C++ using the OpenCV library. This module takes the sequence of images acquired by the capture module as an input and generates a pseudocylindrical equal area projection. This type of projection has the property of keeping undistorted the area of the object that is being projected and therefore is an appropriate projection for the task carried out in this project.

In order to generate this type of projection it is necessary to acquire accurate morphological information of the object under analysis. In this project the morphological information is acquired through edge detection algorithms.

The processing module detects the outline of the object under analysis in every image of the sequence and (according to this data) maps the color information obtaining a bidimensional projection of the object being analyzed. Since fruits have a double-curved surface, the effects of perspective must be compensated using interpolation when mapping the color information.

The analysis module is the tool through which the user of the system takes measures on the pseudocylindrical projection generated before. The first measure taken is the total area. This is an essential value because the system provides measures of the area of regions of interest relative to the whole surface of the piece. Once this measure is taken, the system shows a rectangular cursor that the user hovers over the projection seeking regions of interest to be measured. For this task, the system displays an expanded version of the region covered by the cursor making it easier to find spots.

Whenever the user places the cursor over a region of the projection the system runs an edge detection routine using a statistically calculated threshold. In the event that the automatic detection result is not exact, the user can manually set the threshold value and accurately adjust the contour to the region to be measured and thus get a precise measure of the region’s surface.

Once all measures are taken, the system generates a report consisting of a table of all the measures taken during the analysis, and an image that makes possible to locate (through an index) each of the measured regions in the projection.