Robert Boschgmbh

Charles P Shelton

Sirajum Munir

SeyyedHessamAldin Mohammadmoradi

Omprakash Gnawali

Niluthpol Chowdhury Mithun

A depth-based object-detection convolutional neural network is disclosed. The depth-based object-detection convolutional neural network described herein incorporates a base network and additional structure. The base network is configured to receive a depth image formatted as RGB image data as input, and compute output data indicative of at least one feature of an object in the RGB image data. The additional structure is configured to receive the output data of the base network as input, and compute predictions of the location of a region in the received depth image that includes the object and of a class of the object as output. An object detection device incorporating the depth-based object-detection convolutional neural network is operable in real time using an embedded GPU.

Real-time object detection using depth sensors

An occupancy sensing based system for custodial service management is disclosed that allows schedule custodial services based on actual usage of a site, such as a restroom, instead of following a static schedule and hence can potentially save a significant amount of labor cost. The system is very accurate, privacy aware, and easy to use. In addition to allowing custodial management to optimize their workforce, the system is capable of detecting theft of supplies and performing inventory management. It can be useful in airports, hospitals, stadiums, theaters, academic buildings, convention centers, and many other dynamic places.

Occupancy sensing system for custodial services management

A method of measuring people flow through a doorway using a low-resolution, low-cost infrared array is disclosed. The method includes reading a sequence of sensor data frames from the infrared array; determining a background temperature frame; determining whether sensor data frames include at least one person using noise filtering techniques; identifying at least one group of adjacent pixels corresponding to the at least one person; determining at least one location corresponding to the at least one person; and determining whether the at least one person is a same person as a person in a previous sensor data frame. Based on the matching of a same person through multiple sensor data frames, direction of movement and entry/exit events can be determined. The entry/exit events can advantageously be used to operate and HVAC system, security system, or other such system more efficiently and effectively.

Growing community of inventors

Monroeville, PA, United States of America

Charles P Shelton