A prototype of a relatively cheap laser-based weeding device was developed and tested on couch grass (Elytrigia repens (L.) Desv. ex Nevski) mixed with tomatoes. Three types of laser were used (0.3 W, 1 W, and 5 W). A neural network was trained to identify the weed plants, and a laser guidance system estimated the coordinates of the weed. An algorithm was developed to estimate the energy necessary to harm the weed plants. We also developed a decision model for the weed control device. The energy required to damage a plant depended on the diameter of the plant which was related to plant length. The 1 W laser was not sufficient to eliminate all weed plants and required too long exposure time. The 5 W laser was more efficient but also harmed the crop if the laser beam became split into two during the weeding process. There were several challenges with the device, which needs to be improved upon. In particular, the time of exposure needs to be reduced significantly. Still, the research showed that it is possible to develop a concept for laser weeding using relatively cheap equipment, which can work in complicated situations where weeds and crop are mixed.
The paper presents a detailed analysis of modern techniques that can be used to track gaze with a webcam. We present a practical implementation of the most popular methods for tracking gaze. Various models of deep neural networks that can be involved in the process of online gaze monitoring are reviewed. We introduce a new eye-tracking approach where the effectiveness of using a deep learning method is significantly increased. Implementation is in Python where its application is demonstrated by controlling interaction with the computer. Specifically, a dual coordinate system is given for controlling the computer with the help of a gaze. The first set of coordinates-the position of the face relative to the computer, is implemented by detecting color from the infrared LED via the OpenCV library. The second set of coordinates-giving gaze position-is obtained via the YOLO (v3) package. A method of labeling the eyes is given, in which 3 objects are used to track gaze (to the left, to the right, and in the center).
The manuscript presented an analysis of the work in the field of eye-tracking over the past ten years in the low-cost filed. We researched in detail the methods, algorithms, and developed hardware. To realization, this task we considered the commercial eye-tracking systems with hardware and software and Free software. Additionally, the manuscript considered advances in the neural network fields for eye-tracking tasks and problems which hold back the development of the low-cost eye-tracking system. special attention in the manuscript is given to recommendations for further research in the field of eye-tracking devices in the low-cost field.
11 February 2021 in Journal of Real-Time Image Processing