Corn pest area detection method combining context perception and multi-scale mixed attention

The invention relates to a corn pest area detection method combining context perception and multi-scale mixed attention. The method comprises the following steps: establishing a training sample set; constructing a corn insect pest area detection model: constructing a basic insect pest area detection model, and improving a feature aggregation network in the basic insect pest area detection model by adopting a multi-scale mixed attention module and a context sensing module to obtain the corn insect pest area detection model; training a corn pest area detection model; and insect pest area detection: inputting a to-be-detected sample into the corn insect pest area detection model, generating positioning information of insect pest areas in the to-be-detected sample, and counting the number of the insect pest areas. According to the invention, through the context sensing module, multi-scale context features are introduced to a small target insect pest area to strengthen the expression ability of target features, th