Table 1 Notations

The object detection task of camera i in time slot j

The local computation size of task before offloading partition point v

The offload data size of task at partition point v

The delay tolerance of high-priority tasks

\(\tau _L\)

The delay tolerance of low-priority tasks

\(O_i^j\)

The priority of task \(S_i^j\)

V

The collection of alternative offloading partition points

\(X_i^j\)

The number of objects included in the detection results of task

The detection result x of task \(S_i^j\)

\(Y_{i,x}^j\)

The category of result \(Z_{i,x}^j\)

\(R_{i,x}^j\)

The detection frame size of result \(Z_{i,x}^j\)

\(P_{i,x,y}^j\)

The probability that result \(Z_{i,x}^j\) belongs to category y

\(A_{i,x}^{j,y}\)

The score for result \(Z_{i,x}^j\) corresponding to a dangerous object when it is determined to belong to category y

The threshold for result \(Z_{i,x}^j\) corresponding to a dangerous object when it is determined to belong to category y

\(f_i^j\)

The proportion of the local computing resources assigned to task \(S_i^j\)

\(r_i^j\)

The data rate of transmitting task \(S_i^j\) to edge server

\(h_j\)

The channel gain in time slot j

\(t_{i,l}^{j,v}\)

The time cost of locally processing task \(S_i^j\) before partition point v

\(t_{i,up}^{j,v}\)

The time cost of offloading the feature data for task \(S_i^j\) at partition point v

\(t_{i}^{j,v}\)

The total time cost for task \(S_i^j\) when offloading partition point v is selected

\(u_{i,v}^{j,H}\)

The time utility for high-priority task \(S_i^j\) when offloading partition

point v is selected

\(u_{i,v}^{j,L}\)

The time utility for low-priority task \(S_i^j\) when offloading partition

point v is selected

\(u_i^j\)

The time utility for a task \(S_i^j\) of a certain priority in any time slot j