Plant Species Classification Using a 3D LIDAR Sensor and Machine Learning

Abstract

Abstract-In the domain of agricultural robotics, one major application is crop scouting, e.g., for the task of weed control. For this task a key enabler is a robust detection and classification of the plant and species. Automatically distinguishing between plant species is a challenging task, because some species look very similar. It is also difficult to translate the symbolic high level description of the appearances and the differences between the plants used by humans, into a formal, computer understandable form. Also it is not possible to reliably detect structures, like leaves and branches in 3D data provided by our sensor. One approach to solve this problem is to learn how to classify the species by using a set of example plants and machine learning methods.In this paper we are introducing a method for distinguishing plant species using a 3D LIDAR sensor and supervised learning. For that we have developed a set of size and rotation invariant features and evaluated experimentally which are the most descriptive ones. Besides these features we have also compared different learning methods using the toolbox Weka. It turned out that the best methods for our application are simple logistic regression functions, support vector machines and neural networks. In our experiments we used six different plant species, typically available at common nurseries, and about 20 examples of each species. In the laboratory we were able to identify over 98% of these plants correctly.

Crop recognition under weedy conditions based on 3D imaging for robotic weed control

“…Stereo vision is the method of reconstructing three‐dimensional (3D) surfaces from images taken from multiple 2D sensors by synthesizing objects from different views, and has been applied to plant recognition . The main challenge in using stereo vision for field‐based agricultural systems is the correspondence searching problems caused by the lack of leaf texture, the complexity of the canopy structure, occlusion, and variation in sunlight conditions …”

Automated crop plant detection based on the fusion of color and depth images for robotic weed control

“…Studies of crop or weed plant detection using range data have been reported. Three types of state-of-the-art range sensors were commonly used in agricultural applications, including stereo-vision, LiDAR, and TOF sensors (Weiss, Biber, Laible, Bohlmann, & Zell, 2010). Stereo-vision extracts the distance between the sensor and objects in the field-of-view using images acquired with multiple cameras and exploits advantages of high image resolution, available color information, and detailed textural information (Kise, Zhang, & Rovira Más, 2005), while challenged by sensitivity to illumination and high computational requirements (Tippetts, Lee, Lillywhite, & Archibald, 2016).…”

Crop Row Detection in Maize for Developing Navigation Algorithms Under Changing Plant Growth Stages

“…The RANSAC is a commonly used algorithm for evaluating plane parameters in noisy point cloud data [12], [13]. But also for row or line estimation in image analysis [1], and 2D-LIDAR data [14], [15].…”