Complete Model for Automatic Object Detection and Localisation on Aerial Images using Convolutional Neural Networks

Božić-Štulic, Dunja; Kruzic, Stanko; Gotovac, S.; Papić, Vladan

doi:10.24138/jcomss.v14i1.441

Cited by 9 publications

(7 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Object detection with deep learning methods in remote sensing images is the area directly related to our work. Some of others relevant work in this area include the following studies by Radovic et al (2017) , Božić-Štulić et al (2018) , and Milioto et al (2017) . Remote sensing data and deep learning methods have been put to other usage, for example, estimating the geolocation of ground images by extracting features from UAV images ( Zhai et al, 2017 ) or detecting vehicles from aerial imagery ( Sommer et al, 2017 ).…”

Section: Related Workmentioning

confidence: 99%

Maize Tassel Detection From UAV Imagery Using Deep Learning

et al. 2021

View full text Add to dashboard Cite

The timing of flowering plays a critical role in determining the productivity of agricultural crops. If the crops flower too early, the crop would mature before the end of the growing season, losing the opportunity to capture and use large amounts of light energy. If the crops flower too late, the crop may be killed by the change of seasons before it is ready to harvest. Maize flowering is one of the most important periods where even small amounts of stress can significantly alter yield. In this work, we developed and compared two methods for automatic tassel detection based on the imagery collected from an unmanned aerial vehicle, using deep learning models. The first approach was a customized framework for tassel detection based on convolutional neural network (TD-CNN). The other method was a state-of-the-art object detection technique of the faster region-based CNN (Faster R-CNN), serving as baseline detection accuracy. The evaluation criteria for tassel detection were customized to correctly reflect the needs of tassel detection in an agricultural setting. Although detecting thin tassels in the aerial imagery is challenging, our results showed promising accuracy: the TD-CNN had an F1 score of 95.9% and the Faster R-CNN had 97.9% F1 score. More CNN-based model structures can be investigated in the future for improved accuracy, speed, and generalizability on aerial-based tassel detection.

show abstract

Section: Related Workmentioning

confidence: 99%

Maize Tassel Detection From UAV Imagery Using Deep Learning

et al. 2021

View full text Add to dashboard Cite

show abstract

“…Some studies related to the paper topic are presented in this section. Several authors define object detection and its significance (Demir 2014, Božić-Štulić et al, 2018, Radović et al, 2017. Some authors deal with object detection in general, while the others point out the detection of particular spatial entities.…”

Section: Related Workmentioning

confidence: 99%

“…Faster R-CNN approach for medium-sized objects was elaborated in paper (Zhang et al, 2016). Authors (Božić-Štulić et al, 2018) used a pre-trained Faster R-CNN model for detection of minor deformations from images obtained by UAV surveying technology. The article (Radović et al, 2017) https://doi.org/10.5194/gi-2021-23 Preprint.…”

Section: Related Workmentioning

confidence: 99%

Architecture of Solution for Panoramic Image Blurring in GIS projects Application

Vasić

Davidović

Radosavljević

et al. 2021

Preprint

View full text Add to dashboard Cite

Abstract. Panoramic images captured using laser scanning technologies, which principally produce point clouds, are readily applicable in colorization of point cloud, detailed visual inspection, road defect detection, spatial entities extraction, diverse maps creation etc. This paper underlines the importance of images in modern surveying technologies and different GIS projects at the same time having regard to their anonymization in accordance with GDPR. Namely, it is a legislative requirement that faces of persons and license plates of vehicles in the collected data are blurred. The objective of this paper is to present a novel architecture of the solution for a particular object blurring. The methodology was tested on four data sets counting 5000, 10 000, 15 000 and 20 000 panoramic images respectively. Percentage of accuracy, i.e. successfully detected and blurred objects of interest, was higher than 97 % for each data set.

show abstract

“…The theory of fuzzy sets was created by Zadeh (1965) as an alternative to classical (binary) logic and pointed to the fact that the use of sharp boundaries may be unsuitable in certain cases in which binary logic fails. Since its inception, fuzzy logic has gradually been applied in many fields of human activity, such as geography (Božić-Štulić, Kruzic, Gotovac, & Papić, 2018), medicine (Das, Chowdhury, & Saha, 2012), economics (Shin, & Wang, 2010) and digital technology (Sagar, & Babu, 2012), while it is also widely used in domestic appliances such as washing machines and microwave ovens, in transport in the operation of traffic lights, and in banking.…”

Section: Introductionmentioning

confidence: 99%

Comparison of the Evaluation of Performance Preconditions in Tennis with the Use of Equal and Expertly Judged Criteria Weights

Zháněl

Holeček

Zderčík

2021

International Journal of Computer Science in Sport

View full text Add to dashboard Cite

Tennis performance is influenced by various factors, among which physical performance factors play an important role. The aim of the study was an analysis of possibilities of the use of Saaty’s method for assessing the level of performance prerequisites and comparing the results obtained using equal weights and various weights. The research on Czech female players (U12; n = 211) was based on the results of the TENDIAG1 test battery (9 items) and the results were processed by FuzzME software and relevant statistical methods (correlation coefficient r, Student´s t-test, effect size index d). The results of Saaty’s method show that the most important athletic performance criteria for tennis coaches are the leg reaction time and the running speed, while the least important are endurance and strength. The evaluation using various criteria weights offers a finer scale for assessing athletes’ performance prerequisites despite the proven high degree of association between the results obtained with equal and various weights and the insignificant difference of mean values. The results have shown possibilities for the use of a fuzzy approach in sports practice and motivate further research towards broadening the structure or the number of evaluation criteria.

show abstract

Complete Model for Automatic Object Detection and Localisation on Aerial Images using Convolutional Neural Networks

Cited by 9 publications

References 17 publications

Maize Tassel Detection From UAV Imagery Using Deep Learning

Maize Tassel Detection From UAV Imagery Using Deep Learning

Architecture of Solution for Panoramic Image Blurring in GIS projects Application

Comparison of the Evaluation of Performance Preconditions in Tennis with the Use of Equal and Expertly Judged Criteria Weights

Contact Info

Product

Resources

About