Vision-based automatic fruit counting with UAV

Hubert Szolc, Mateusz Wasala, Remigiusz Mietla, Kacper Iwicki, Tomasz Kryjak · 2025

Read Vision-based automatic fruit counting with UAV by Hubert Szolc, Mateusz Wasala, Remigiusz Mietla, Kacper Iwicki, Tomasz Kryjak (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-13080.txt.

Introduction

Hubert Szolc and co-authors describe in 2025 a drone system that counts fruit on its own for smart farming. A camera and a depth sensor feed a vision method that finds the fruit as the drone flies a planned route.

The system also plans flight paths that keep time and distance low. In test flights run in a simulator, the method scored well across five hundred missions. The team then flew it at a drone contest in 2024, where it placed sixth of twenty three teams and reached the finals. A reader sees a full loop from flight plan to fruit count.

Catalog record

CREATORHubert Szolc, Mateusz Wasala, Remigiusz Mietla, Kacper Iwicki, Tomasz Kryjak
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE22 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-13080

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Vision-based automatic fruit counting with UAV Hubert Szolc, Mateusz Wasala, Remigiusz Mietla, Kacper Iwicki, Tomasz Kryjak Abstract The use of unmanned aerial vehicles (UAVs) for smart agriculture is becoming increasingly popular. This is evidenced by recent scientific works, as well as the various competitions organised on this topic. Therefore, in this work we present a system for automatic fruit counting using UAVs. To detect them, our solution uses a vision algorithm that processes streams from an RGB camera and a depth sensor using classical image operations. Our system also allows the planning and execution of flight trajectories, taking into account the minimisation of flight time and distance covered. We tested the proposed solution in simulation and obtained an average score of 87.27/100 points from a total of 500 missions. We also submitted it to the UAV Competition organised as part of the ICUAS 2024 conference, where we achieved an average score of 84.83/100 points, placing 6th in a field of 23 teams and advancing to the finals. keywords object counting, unmanned aerial vehicles, UAV, camera, smart agriculture 1 Introduction Unmanned aerial vehicles (UAVs, also known as drones) have recently gained a significant interest from both scientific and commercial communities. This is due to the relatively wide range of applications, including military missions, cargo transport, terrain inspection, environmental mapping and spectacular cinematography. One interesting area where UAVs can bring tangible benefits is in smart agriculture Srivastava and Prakash (2023) , particularly through the use of image processing algorithms Su et al. (2023) . The potential impact of UAVs in smart agriculture is confirmed, among other things, by recent competitions dedicated to this topic. One of the most relevant examples is the Unmanned Aerial Vehicle (UAV) Competition accompanying the International Conference on Unmanned Aircraft Systems (ICUAS) 2024. The challenge was to count different types of fruit using a quadcopter equipped with an RGB camera and a depth sensor. For this purpose, the organisers prepared a special simulation environment Markovic et al. (2023) to prototype and evaluate solutions. Figure 1: The overview of our vision-based system for automatic fruit counting with UAV. Proposed solution uses a streams from an RGB camera and a depth sensor and process them with classical image processing algorithms. Additionally, our system enables planning and execution of flight trajectories, optimizing for minimal flight time and distance covered. Automatic fruit counting is an important task for several reasons. First of all, it facilitates the new branch of agriculture called smart agriculture . Thanks to fruit counting, it is possible to monitor yields more accurately, to predict the size of the harvest in order to optimise the logistics and sales of the collection, and to detect early fruit drop or uneven ripening. It also helps to monitor the health of the crop, for example identifying nutrient deficiencies by analysing the distribution and size of the fruit. For large orchards, fruit counting provides a continuous update on the quantity of fruit, which is impractical to achieve manually. An automated system to perform this task is therefore essential, especially in large-scale agriculture. In this paper, we present our solution to the aforementioned competition task. For fruit detection, we propose a novel vision system that uses classical video stream processing algorithms. Thus, it provides a relatively short processing time to count objects while the UAV is flying. For route planning, we use the Fast-TSP algorithm Mulvad (2022) . It ensures minimisation of flight time and distance covered, which is associated with reduced energy consumption. The fruit counting system that we have prepared is shown schematically in Figure 1 . Our solution enabled us to reach the 6th place (out of 23 participants) in the classification of the simulation phase of the competition. The main contribution of this paper can be summarised as follows: • We have developed a proprietary fruit detection algorithm which, together with the control strategy, forms a system for automatic fruit counting using an unmanned aerial vehicle. • We have tested our approach in a simulation environment according to the rules of UAV Competition at ICUAS 2024. We make our entire code available under an open licence in the Github repository 1 1 1 Removed for blind revision . The remainder of this paper is organised as follows. In Section 2 we describe similar approaches to automatic object counting with the use of unmanned aerial vehicles. Our own solution to this problem is presented in Section 3 . In Section 4 we present results obtained with our method along with a discussion. The paper concludes with a summary in Section 5 , in which we also outline plans for further work. 2 Related work The use of unmanned vehicles in smart agriculture is growing rapidly, offering new opportunities for crop monitoring, plant health assessment and yield counting. These methods are gaining popularity due to their efficiency, high accuracy and ability to perform large-scale analysis at relatively low cost. The paper Lukas Meyer (2024) presents FruitNeRF, an innovative fruit counting system using Neural Radiance Fields (NeRF). It enables fruit detection and classification based on different images of the same plant. The system combines advanced image segmentation (e.g. SAM and U-Net) with the NeRF method to create 3D point clouds representing the fruit. In addition, the counting process uses cascade clustering, which accurately identifies individual fruits and their clusters based on spatial volume. Tests were carried out on synthetic data (e.g. apples, plums, mangoes) and real data (apple trees in an orchard), achieving high accuracy (F1 score: up to 0.95 for reference masks). In both cases, the input data was collected by manual camera positioning without the use of drones. The paper Liu et al. (2019) proposes a fruit counting system using only a single camera mounted on an unmanned ground vehicle (UGV). A Faster-RCNN network is used to detect fruits and tree trunks. They are then tracked in subsequent images using a Kalman filter and optical flow estimation. A key part of the solution is the transformation of object tracks from the 2D plane to 3D landmarks. This eliminates the problem of double fruit counting. The solution, tested on mango fruit data, achieved results comparable to more sophisticated systems that also use other sensors. The paper Kumar and Behera (2024a) presents a comprehensive Drone-Bee system for autonomous fruit picking in environments without access to a GNSS signal, such as vertical farms. The authors developed a lightweight grapple (to be mounted on a UAV), a precision localisation system using SLAM, vision-based perception and a drone flight control algorithm. The system was optimised to take into account the limited computational power of on-board devices and enables precise fruit picking in challenging conditions such as external disturbances (e.g. wind) or restricted workspaces. The Drone-Bee uses the proprietary Fast Fruit Detector (FFD) algorithm Kumar and Behera (2024b) , which is based on the Detection Transformer (DETR) model and was implemented on the Nvidia Jetson NX platform. The cited work shows that there is currently considerable interest in the subject of object counting for smart agriculture. This is particularly true for the fruit detection and localisation algorithms, which are dominated by approaches based on image processing algorithms. In one of the cited papers, the authors presented a

Read the complete text