We are a team of researchers working on a robotic project and recently bought multiple 5MP OV5647 Cameras witch CS-mount + lens (LS-2717CS) (https://www.arducam.com/product/arducam-5mp-cs-mount-camera-module-raspbery-pi/). We are very happy with the quality and sharpness of the image and started assembling multiple robots using the camera. We, however, noticed that each camera had a noticeably irregular mounting angle. We are using them so that the sensor is approximately 190mm from a plan (a flat object) we take a picture of. For instance, if we take several images using exactly the same rig, but different cameras, we observe the centre of the image can shift by more than 200 px (on a 2592 px wide image). This is an issue in so far as , since the axis of the camera is not always orthogonal to the plane of the sensor, one edge of our images may end-up out of focus and off-centred. We went further and systematically tested the offset angle using two boards, two mount adaptors and tree lenses. We also played with rotated the mount adaptor 180 degrees as it is symmetrical. It looks as though certain combinations of lense and mount have a widely large angular error. For instance, the centre may be off by more than 10mm from a 190mm high sensor, which is more than 3 degrees. My feeling so far is that it seems to be due to imperfections in how the mount adaptor is machined. We have several questions. 1. Is such a large variation between products expected. 2. Is there an efficient and reliable way to improve this?
Supporting material below:
The angle is off in this camera, making the right margin out of focus and the centre off:
Rig to test focus:
Example of two pictures taken with the same lense, just swapping the mount. notice how much the central point has shifted:
Results of systematic tests (x and y are the position of the central tip in a 2592 x 1944 picamera image):