A Vision-Based Bee Counting Algorithm for Electronic Monitoring of Langsthroth Beehives

An algorithm is presented to count bee numbers in images of Langsthroth hive entrances. The algorithm computes approximate bee counts by adjusting the brightness of the image, cropping a white or green area in the image, removing the background and noise from the cropped area, finding the total number of bee pixels, and dividing that number by the average number of pixels in a single bee. On 1005 images with green landing pads, the algorithm achieved an accuracy of 80 percent when compared to the human bee counting. On 776 images with white landing pads, the algorithm achieved an accuracy of 85% compared to the human bee counting