Basically, a collision between a stationary and non-stationary object would more or less necessitate the same formula. Assume Vehicle 1 is the vehicle in motion, and initiating contact. Also assume that both vehicles, if moving, are doing so on the exact same vector.
Change in velocity = Velocity of vehicle 2 - Velocity of vehicle 1.
∆V=(V2-V1)
Assume ∆V is transferred to the loose object in your vehicle as its new velocity, vectorized accordingly to the direction of impact. (as we have assumed both vehicles are moving along the same vector, this is negligible, but in reality, some velocity is lost here, as well in the force lost in the crumpling of metal between vehicles) At this point, assuming the vectorized force is enough to break the inertia of the loose object (in reality, some velocity is lost in this transfer), and assuming there is a clear path for the loose object to follow along the force of influence (which in an object loose on the floorboard is unlikely):
F=((M*V)/∆T)
Force = (Mass x Velocity)/change in time
Therefore, the force of the impact of the loose object is equal to its mass multiplied by its velocity, over the change in time.
So, if you have a 1 pound object, which is traveling at 1 foot/sec, the force of said object would be equal to 1 foot/lb. (the same units used in measuring torque)
However, impact is not measured in such terms. In order to calculate the force/surface area of the impact you would need the variable of surface area impact. I.E. A knife's edge, with the same amount of force behind it, would generate a much higher P.S.I (pounds per square inch) than would a bowling ball.
In conclusion, physics can't really help scare a knowledgeable person into believing the umbrella on the floorboard would turn into a lethal projectile in any/most/all collisions, as too many variables are unknown. (How much is not transferred between vehicles due to not hitting squarely, how much force is used in the crumpling of vehicles, how much is lost in inertia, how much is lost in ricochets, and finally angle of impact of loose object and surface area of impact) While the possibility exists of a poorly mounted child seat breaking loose and lethally striking the passenger in front, there are too many variables to plug into a simple java application and get a regurgitated answer.
Theoretically, assuming all impacts convey a perfect transfer of energy, you could employ such tactics. However, we don't live in the 1950s where vehicles were designed to survive collisions regardless of how many humans were converted to so much ground meat, and safety features designed into the very sheetmetal and subframes protect us. To imply all modern collisions result in perfect transfer of energy to the objects and people within the vehicles is misleading and irresponsible.
Crash forces between vehicles are indeed on a massive scale. Tons of metal crashing into tons of metal at high rates of speed. However, the question posed here is not about the force of impact between vehicles, but rather the force of objects in said vehicles on other surfaces in the same environment. These by their very nature would tend to be much less dramatic, and the only constant would be the velocity of the collisions, albeit diluted.