UAV (Unmanned Aerial Vehicle) and UGV (Unmanned Ground Vehicle) Computer Vision solutions

Computer Vision solutions consist of hardware and software. From the hardware side, it is important to consider the requirements to make particular analysis possible (in real time or not).

Real time and close to real time analysis requires specific hardware with specific parallel computing capabilities. Modern solutions often use GPU hardware for that.

It depends largely on the analysis type and optimization but for reference one modern average GPU card (11 TFLOPS) can run modern object detection algorithm on Full HD video resolution 30FPS.

UGV (Unmanned Ground Vehicle) usually has 4 on-board cameras to cover 360 degrees. With the same frame-rate for all the cameras, the on-board processing capability would have to be roughly 44 TFLOPS.

Computer Vision Systems for PPE work safety and for industrial safety

AI based video analyzer (CAFALYZER) for industrial and robotics safety.

CAFALYZER provides computer vision for automated situation assessment:
1. Inspection of the workers wearing PPE (Personal Protective Equipment).
2. Checking machine tools when protective sensors are covered.
3. Checking the safety of robots.

Adjustment of computer vision alert thresholds and performance of specific functions is possible when a work safety coordinator changes settings of CAFALYZER.

Computer Vision artificial neural network weights are calculated by dedicated hardware and software on a workstation but it requires a large dataset of specifically prepared images for that. Raw images can be aquired directly from the stationary cameras with some coordination from a work safety expert and AI expert taking reference images from actual workers and machinery.

CVS to evaluate the efficiency of processes

  1. Analyze the trajectories, speeds and delays of mobile objects and personnel to find optimization options by changing the layout of the work area or changing the timing.
  2. Analyze congestions due to less than perfect ad-hoc storage areas.
  3. Analyze the safety regarding storage use.
  4. Find time-critical logistics problems that influence the overall production efficiency the most.
  5. Analyze trends (cluttering, contamination of surfaces, build up of storage area).

CVS to control quality of products

Visual inspection by human(s) is not fast nor reliable enough by modern standards. As long as products can have known types of defects, it is common to use dedicated machine vision apparatus for detecting them automatically. When this quality control apparatus detects a defective product, it can quickly send command to an actuator, which removes the product from the production line to a container that can later be inspected by humans with no rush.

High level workflow Computer Vision System (CVS)