סמינר מחלקה של עדי כהן - הערכת אוריינטציה של עלי עגבנייה באמצעות למידה עמוקה לשם זיהוי מוקדם של מחלות

30 בנובמבר 2022, 14:00 - 15:00 
פקולטה להנדסה 
0
סמינר מחלקה של עדי כהן - הערכת אוריינטציה של עלי עגבנייה באמצעות למידה עמוקה לשם זיהוי מוקדם של מחלות

 

 

 

School of Mechanical Engineering Seminar

Wednesday, November 30 2022 at 14:00
Wolfson Building of Mechanical Engineering, Room 206

 

Estimation of Tomato Leaves Orientation for Early Detection of Diseases Using Deep Neural Network Model

Adi Cohen

MSc Avishay Sintov

 

Early detection of diseases in tomato is essential to prevent infection of fruit and other plant parts, spreading of the disease in the plot and to minimize crop losses and reduction of quality. A robotic system equipped with an electric resistance sensor is developed to reach, grasp, and measure the conductivity of a tomato leaf. The electric resistance seems to be a significant indicator regarding the plant's condition in early stages of the disease. To grasp a leaf and perform the measurement with a robotic arm, a method for automatically identifying the leaf orientation was developed using data from images of leaves as input to a Deep Neural-Network (ANN) model. An experiment was conducted using the UR5 robotic arm to collect and label leaf images in different orientations and distances. Each image was processed to extract an array containing a list of boundary points of the leaf in every image as well as information of the robotic position and orientation in order to prepare an input for the ANN. The ANN then received this input and produces a spatial vector normal to the leaf's surface. To determine the leaf orientation, several analyses were conducted using one, two and three images of the same leaf in different poses and the changes in the manipulator orientation to take them. The main results show an average accuracy of 11.5 degrees between predicted and real orientation of the leaf's surface which in most cases is sufficient to grasp the leaf and perform the measurement. In addition, the ANN model results were analyzed to optimize the robotic motion to best predict the leaf's orientation and improve the accuracy. Those results are satisfying for the robotic end effector to perform the measurement of the leaf's resistance.

 

Join Zoom Meetin https://us02web.zoom.us/j/82108132163?pwd=Z2h4UzNzUS9mbXplT0lMU1pZenFEQT09

 

אוניברסיטת תל אביב עושה כל מאמץ לכבד זכויות יוצרים. אם בבעלותך זכויות יוצרים בתכנים שנמצאים פה ו/או השימוש
שנעשה בתכנים אלה לדעתך מפר זכויות, נא לפנות בהקדם לכתובת שכאן >>