Bulldozer blade and ripper shape as a manufacturer tell in satellite imagery

A dozer sitting on a pad at 30cm resolution gives you a count. It does not hand you a make. But the blade and the ripper behind it carry more brand information than most analysts give them credit for, if you know what to look at and what to ignore.
What the blade silhouette actually tells you
Blade shape sorts into a handful of recognizable families before it sorts into brands: straight (S-blade), universal (U-blade), semi-universal (SU-blade), and angle blades with the trunnion mount visible on one side. From overhead, you're reading the blade's curvature, its end bit wings, and how far it extends past the track frame on each side.
That family read is reliable. A U-blade's wing extensions throw a distinct shadow pair that a straight blade never produces, even at dusk sun angles where shadow length is doing most of the work for you. What's harder is going from "this is a U-blade" to "this is a Cat D8 versus a Komatsu D155." OEMs converge on similar blade geometry within a size class because the engineering constraints are the same: push volume, ground clearance, c-frame mounting points. A large dozer's U-blade from one manufacturer can look close enough to another's that silhouette alone won't separate them without a reference chip at a known scale.
Where it gets more useful is the trunnion and push-arm mounting pattern, which does vary by OEM and stays visible even when paint color is washed out by overexposed imagery or a dust haze. Push-arm attachment points sit at consistent offsets for a given OEM's frame design, and that offset is measurable in a VHR chip if you've got pixel resolution to spare.
Ripper shank configuration as a secondary signal
The ripper is the better tell, honestly. Single-shank versus multi-shank (usually three-shank) configurations are a design choice that tracks fairly closely with OEM model lines and the ground conditions a given fleet is built for. A three-shank ripper frame reads as a distinct rectangular mass behind the main frame, wider and flatter than a single-shank parallelogram linkage. At native VHR resolution that width difference is one of the more stable shape reads you'll get off a dozer, because unlike the blade it isn't constantly repositioned relative to the ground.
Shank spacing and the shape of the ripper beam itself also vary enough between manufacturers that a trained eye, or a model trained on enough reference chips, can narrow the candidate list even when the blade read is ambiguous. Some operators run without a ripper at all, which removes the signal entirely rather than giving you a negative one, so don't read "no ripper visible" as evidence of anything beyond "no ripper visible."
Why this is an attribution input, not an attribution answer
None of this works in isolation. Blade family narrows you to a handful of OEMs that build that geometry at that size class. Ripper configuration narrows further. Counterweight shape and paint color, when the imagery resolves color at all, do the rest of the work. Treat blade and ripper shape as one input into a fleet-share read, not a standalone identification method, and you'll avoid the mistake of overclaiming a single-unit match off shadow geometry alone.
This is also exactly where a fleet-share read differs from a simple equipment count. Counting dozers on a competitor's site tells you how many machines are working. Attributing each one to a manufacturer, using blade family, ripper configuration, counterweight silhouette and whatever color signal the pass gives you, is a harder problem and a different deliverable. Machine Brand Recognition is built around that second problem specifically, reading the combination of those cues across a detected fleet rather than flagging any one shape in isolation.
If brand-level fleet share on a job site or yard is the question you're actually trying to answer, that's the use case worth getting on the early access list for.