I like your discussion and you are certainly not alone in this. If is a difficult area and interests many so here is my input in this. If it helps or not … ;).
Motion detection is a tricky business and is in focus for research and development since many years. As a lot of high tech stuff today used in everyday systems and applications it emanates from the military. Search the Internet and you will find a lot of very interesting articles about this and the development of mathematical algorithms trying to find the target in different ways. Today we use only a small part of this in the free or licensed softwares, but still the purpose is the same; to keep track of the mail man, check on our pets, children, unwanted guests in the backyard, …
In NCS there are a couple of different algorithms to chose from, but the basics are the two parameters Motion Sensibility (not sensitivity) and Threshold. Here you can find more about these two parameters Difference between sensibility and threshold
-Sensibility is a parameter connected to how the motion algorithm should respond to a fast or slow movement of the object.
-Threshold is a parameter connected to how the motion algorithm should react to how much of the sensor area in the camera that the object cover.
Since this is about motion detection these two parameters closely interacts to trigger the system to start a recording. It is important to know that in these simplified systems the motion detection system do not know if what generates the trigger for a recording is the arriving mail man or a cloud in front of the the sun that generates a moving shadow on the ground. Of cause there are systems that can distinguish between these two things. More sophisticated software and hardware with a different price tag, but this also put more demands of the user. There are many articles about this and here is two of them that are quite enlightening:
-Motion detection and objects tracking algorithm implementation
-Motion Detection Algorithms
So, what to do?
Finding the target in an outside environment is difficult if you want no “false alarms”. If you know what you are looking for, like a person, I suggest that a person pass in front of the camera on the desired distance and then you change the Threshold and Sensibility until you get the best result. Most likely it is the Threshold that you find easy to adjust. Remember, that the person will cover different amount of area on the camera sensor depending on the distance from the camera so the Threshold will also determine at what distance the moving person will be detected.
I would set the Sensibility to Normal since that parameter is more of a fine tuning. Maybe a very slowly moving Turtle is not of interest to record, but it still covers a large area on the sensor. Then it can be of interest to decrease the Sensibility.
If it is difficult to find out what is actually triggering the system I would set it to continuous recording and on motion detection for a day. Then it is possible to see everything and also the moment when the system trigger. If it is outside it can be a shadow, leaves on a tree or what ever is in the camera field of view.
I also like to use the View -> Motion detection since that assist me a lot to see what trigger the system. I have been testing different settings and have not found a way to also record that view. I certainly agree with you on this so I am sending this to our developers if thay can do their magic and involve this overlay also in the recorded file.
During night time it is usually ever more difficult since there are a lot of annoying bugs, maybe snow flakes, rain drops, … that reflects the light from the IR LEDs back into the camera and generates a lot of false recordings. To solve this one solution is to mount an external IR source a bit from the camera and turn off The IR LEDs in the camera.
Good luck guys and you know where to find me ;).