Automatic aphid counting based on deep learning and robotics technology in unstructured agriculture environments

Aims:

Aphids can cause direct damage and indirect virus transmission to crops. Timely monitoring the number of aphids can prevent the large-scale outbreak of aphids. However, the manual counting of aphids which is commonly used at present is inefficient, and it requires professional staff to complete it. Therefore, in this project, we will develop an automated aphid counting method combined with the agricultural robot and deep learning, so as to replace the manual aphid counting.

Objectives:

1. The robot can automatically navigate to the location of the yellow water pan traps device that collects aphids in the farmland and obtain images containing aphids. 2. Designing an aphid counting CNN to automatically count the number of aphids in the image.

3. The robot can remotely transmit the aphid counting results and the position coordinates of the corresponding yellow water pan trap to the customer service terminal. Impacts:

Compared with the manual counting method, the developed automatic aphid counting method for agricultural robots will greatly improve the efficiency of aphid counting. At the same time, the designed automatic aphid counting CNN can be widely used in the identification and counting of agricultural small pests similar to aphids.