In a preferred embodiment, the cameras 18₁ . . . 18_n are configured as code readers in which the control and evaluation unit 26 additionally acts as a decoding unit for reading code information and corresponding pre-processing for localizing and preparing code regions. It is also conceivable to detect streams of objects 14 without codes 16 and accordingly to dispense with the decoding unit itself or its use, for example for an inspection or quality control. In addition, a decoding can also take place downstream outside the cameras 18₁ . . . 18_n on their image data, in particular in the higher ranking control 34.

The conveyor belt 12 is too wide to be detected by an individual camera 18₁ . . . 18_n. The fields of view 20₁ . . . 20_n of the cameras 18₁ . . . 18_n therefore preferably complement one another in the transverse direction of the conveyor belt 12 with a partial mutual overlap. An optional geometry detection sensor 36, for example in the form of a distance measuring laser scanner known per se, is located above the individual cameras 18₁ . . . 18_n with respect to the direction of movement of the conveyor belt 12 to measure the three-dimensional contour of the objects 14 on the conveyor belt 12 in advance, with dimensions, volume, position, and shape resulting therefrom. In addition, the geometry data can be used to trigger the recordings and to focus the cameras 18₁ . . . 18_n.