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

• DOI: 10.3390/s20072126
• Publication Date: 2020-04-13T14:41:52Z
• Authors: Jayme Garcia Arna...; Luciano Vieira Ko...; Patrícia Menezes ...; Andrea Roberto Bu...
• Keywords: 2. Zero hunger; Aircraft; Chemical technology; TP1-1185; 02 engineering and technology; Nelore breed; Canchim breed; Article; convolutional neural networks; Image Processing, Computer-Assisted; 0202 electrical engineering, electronic engineering, information engineering; Animals; mathematical morphology; Cattle; Neural Networks, Computer; unmanned aerial vehicles

The management of livestock in extensive production systems may be challenging, especially large areas. Using Unmanned Aerial Vehicles (UAVs) to collect images from the area interest is quickly becoming a viable alternative, but suitable algorithms for extraction relevant information are still rare. This article proposes method counting cattle which combines deep learning model rough animal location, color space manipulation increase contrast between animals and background, mathematical morphology isolate infer number individuals clustered groups, image matching take into account overlap. Nelore Canchim breeds as case study, proposed approach yields accuracies over 90% under wide variety conditions backgrounds.