Bees Detection on Images: Study of Different Color Models for Neural Networks

Dembski, Jerzy; Szymański, Julian

doi:10.1007/978-3-030-05366-6_25

Cited by 4 publications

(2 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…According to previous research, RGB is more suitable for neural network learning compared to the one-dimensional value according to the H values of RGB and HSV [13]. A color model refers to 3D array data for expressing colors, and each dimension has a component value for implementing the color.…”

Section: Color Model Conversionmentioning

confidence: 99%

Identifying an Image-Processing Method for Detection of Bee Mite in Honey Bee Based on Keypoint Analysis

Lee,

Kim,

Kim

et al. 2023

Agriculture

View full text Add to dashboard Cite

Economic and ecosystem issues associated with beekeeping may stem from bee mites rather than other bee diseases. The honey mites that stick to bees are small and possess a reddish-brown color, rendering it difficult to distinguish them with the naked eye. Objective and rapid technologies to detect bee mites are required. Image processing considerably improves detection performance. Therefore, this study proposes an image-processing method that can increase the detection performance of bee mites. A keypoint detection algorithm was implemented to identify keypoint location and frequencies in images of bees and bee mites. These parameters were analyzed to determine the rational measurement distance and image-processing. The change in the number of keypoints was analyzed by applying five-color model conversion, histogram normalization, and two-histogram equalization. The performance of the keypoints was verified by matching images with infested bees and mites. Among 30 given cases of image processing, the method applying normalization and equalization in the RGB color model image produced consistent quality data and was the most valid keypoint. Optimal image processing worked effectively in the measured 300 mm data in the range 300–1100 mm. The results of this study show that diverse image-processing techniques help to enhance the quality of bee mite detection significantly. This approach can be used in conjunction with an object detection deep-learning algorithm to monitor bee mites and diseases.

show abstract

Section: Color Model Conversionmentioning

confidence: 99%

Identifying an Image-Processing Method for Detection of Bee Mite in Honey Bee Based on Keypoint Analysis

Lee,

Kim,

Kim

et al. 2023

Agriculture

View full text Add to dashboard Cite

show abstract

“…There are multiple studies about image processing in the agricultural field for classification and detection of marks on fruit like apple, pear, etc., determination of the placement of fruit on branches, and detection of the dimensions and quality of agricultural products (Dubey & Jalal 2016;Capizzi et al 2016;Yang et al 2017;Zhang et al 2017;Ponce Aquino & Andújar 2019;Songaet al 2020). In the apiculture field, there are image processing studies related to counting bees, determining the gender of bees and observing bee movements (Dembski & Szymański 2019;Ngoa et al 2019). Additionally, there are studies related to classification of honeycomb cells, detection of size and detection of only large larvae present in the cells in the literature (Sparavigna 2016; Alves et al 2020;Giraud 2020).…”

Section: Introductionmentioning

confidence: 99%

Development of AI Based Larvae Transfer Machine for Royal Jelly Production

Güneş

2023

J Agr Sci-Tarim Bili

View full text Add to dashboard Cite

Honeybees produce many different products beneficial to humans. One of these of is royal jelly which is the bee product with highest nutritional value but is most difficult to produce. The most time-consuming procedure in royal jelly production involves removing larvae with ideal size from the honeycomb cells and transferring them to queen cups. In order to increase the speed of the larva transfer process and perform it without labor power, a machine autonomically performing larva transfer was developed in three stages. Firstly, a CNC platform that can move on three axes above the honeycomb was created. In the second stage, a camera device was developed to image the larvae and mounted on the platform. Later larvae were photographed with this device and labelled. Tagged photos have been quadrupled by data augmentation methods. A Mobiledet+SSDLite deep learning model was trained with these photographs and this model identified larvae with ideal size with 96% success. Additionally, the central points of the honeycomb cells were identified with the Hough circles method. In the third and final stage, a device which can transfer the identified larvae from the honeycomb cells to the queen cups was developed and mounted on the platform. Later general software controlling the platform and devices was developed. At the end of this study, for the first time in the literature, an artificial intelligence-supported machine was developed for automatic transfer of ideal larvae from natural honeycombs for royal jelly production.

show abstract

Weighted Clustering for Bees Detection on Video Images

Dembski

Szymański

2020

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Bees Detection on Images: Study of Different Color Models for Neural Networks

Cited by 4 publications

References 19 publications

Identifying an Image-Processing Method for Detection of Bee Mite in Honey Bee Based on Keypoint Analysis

Identifying an Image-Processing Method for Detection of Bee Mite in Honey Bee Based on Keypoint Analysis

Development of AI Based Larvae Transfer Machine for Royal Jelly Production

Weighted Clustering for Bees Detection on Video Images

Contact Info

Product

Resources

About