Detectors

Detectors

Detectors are the algorithms responsible for interpreting the input from the sensors and create Obstacle objects that can be interpreted by the avoidance strategies.

The project provides two Detectors. Note that the prefix indicates the type of sensor the detector works with.

• Depth Image Obstacle Detector
• Depth Image Polar Histogram Detector

Note that most detectors use the concept of readings density. This is used to reduce the effect cause by noise in the sensor input.

Depth Image Obstacle Detector

This detector will use the concept of reading blobs to find obstacles in the sensor input. Given the input data, this detector will group each reading into labels. For each reading, the detector will check its left, top-left, top and top-right neighbors. If they are part of a label, the new reading will be added to the same label. If not, and if the current reading is bellow the safe distance threshold, a new label will be created. Once all the readings have been processed, the detector will remove the labels that have a readings count bellow a provided threshold. Then, For the remaining labels, an obstacle will be created centered at the blob's center and with the distance to the vehicle given by the smallest reading in that group.

Depth Image Polar Histogram Detector

This detector expects that the vehicle is going forward in the direction where the sensor is pointed. Given the sensor input, the detector will check the middle slice of the depth map and will divide it in multiple slices. A provided angle step defines the size of each slice. For each slice, the detector will check the density of readings that are bellow a provided distance threshold. If the minimum density is met, the detector will create an obstacle centered in that slice center, and its distance from the vehicle will be the minimum distance reading identified in that slice.