Feature Maps

A feature map is the output produced by applying a kernel across positions.

If a kernel acts like an edge detector, the feature map records where that detector responds strongly. If another kernel acts like a texture detector, its feature map records texture-like regions.

The phrase "feature map" is literal enough:

feature: what the kernel detects
map: where it appears

A convolutional layer usually has many kernels, so it produces many feature maps. Later layers can combine them.