Table 1. Detailed steps for the object recognition and trapping.

Step 1	Initialize Image , t = 0.
Step 2	Capture an image containing the nanoknife and take it as the template . Then, calculate and initialize the position of template T: .
Step 3	Extract the nanoknife tip coordinate in the Template T: First, extract edge of nanoknife using Canny edge detection algorithm based on maximum local gradient value and judgement rules of adaptive histogram threshold; Second, extract all impossible corners of nanoknife using CSS detection algorithm in variable scale space along the above edge curve. Third, get nanoknife tip corner A in the template T:
Step 4	Recognize the nanoknife’s tip position Pt at time t: First, search in It from the last position , and match the template T to get the new matching position ; Second, calculate the nanoknife’s tip position Pt in It. Since the nanoknife tip corner A in the template is constant, its position Pt in It can be represented by Bt + A, which is .
Step 5	Calculate the distance between nanoknife and target cell. The sample position S (Sx, Sy) can be obtained via the human-machine interface. So the distance Dt(Dx, Dy) can be represented by: , .
Step 6	Reduce search region in the original image, and update .
Step 7	If Dx < ε0 and Dy < ε0, where ε0 is an arbitrary small positive number, the recognition procedure stops; otherwise set t = t + 1 and go to step 4.