A Study on the Detection of Cattle in UAV Images Using Deep Learning

Abstract

Unmanned aerial vehicles (UAVs) are being increasingly viewed as valuable tools to aid the management of farms. This kind of technology can be particularly useful in the context of extensive cattle farming, as production areas tend to be expansive and animals tend to be more loosely monitored. With the advent of deep learning, and convolutional neural networks (CNNs) in particular, extracting relevant information from aerial images has become more effective. Despite the technological advancements in drone, imaging and machine learning technologies, the application of UAVs for cattle monitoring is far from being thoroughly studied, with many research gaps still remaining. In this context, the objectives of this study were threefold: (1) to determine the highest possible accuracy that could be achieved in the detection of animals of the Canchim breed, which is visually similar to the Nelore breed (Bos taurus indicus); (2) to determine the ideal ground sample distance (GSD) for animal detection; (3) to determine the most accurate CNN architecture for this specific problem. The experiments involved 1853 images containing 8629 samples of animals, and 15 different CNN architectures were tested. A total of 900 models were trained (15 CNN architectures × 3 spacial resolutions × 2 datasets × 10-fold cross validation), allowing for a deep analysis of the several aspects that impact the detection of cattle using aerial images captured using UAVs. Results revealed that many CNN architectures are robust enough to reliably detect animals in aerial images even under far from ideal conditions, indicating the viability of using UAVs for cattle monitoring.

Counting Cattle in UAV Images—Dealing with Clustered Animals and Animal/Background Contrast Changes

“…The first approach, semantic segmentation, associates each pixel in the image to a class label. Although there are several deep learning architectures capable of performing semantic segmentation (e.g., U-Net [10], DeepLab [11] and FastFCN [12]), this kind of technique has not yet been applied to cattle recognition and counting, probably due to the difficulty involved in the annotation of the reference data [13]. The second approach, object detection, delineates a box bounding the objects of interest.…”

“…This article proposes a four-step approach combining the inference capabilities of deep learning models with finely tuned image processing for extraction of relevant information. In the first step, which was described in detail in a previous work [13], each image is divided into squares using a regular grid, and squares containing parts of animals are identified by means of a CNN. In the second step, color space manipulations are used to enhance contrast between animals and background and thresholds are applied to generate binary masks.…”

“…It is worth noting that although the CNN models used in the algorithm successfully deal with poor contrast and illumination [13], they only provide a rough estimation for the position of the animals. Determining how many animals are actually present in the selected regions is challenging under these conditions due to difficulties in animal delineation.…”

“…For this reason, the proposed algorithm has many steps specifically designed to deal with each problematic situation. It is also important to highlight that the CNN models described in Reference [13] were employed without modification, that is, there were no new experiments regarding the deep learning part of the algorithm.…”

Scheduling of a Parcel Delivery System Consisting of an Aerial Drone Interacting with Public Transportation Vehicles

“…With the fast development of drones (also called unmanned aerial vehicles (UAVs)) technology, drones have found various applications in civilian domains [1], such as wireless communication support [2], structural health inspection [3], farming [4], surveillance and monitoring [5][6][7] and parcel delivery [8,9]. Thanks to their mobility and flexibility, many logistics companies and many advanced control approaches such as autonomous landing [10], such as Amazon, Alibaba, DHL, SF Express, etc., have started to pay more attention to the application of drones in parcel delivery.…”